IOT APPLICATIONS FOR CO2 AND SOIL MOISTURE MEASUREMENT

CHAPTER 1

1. INTRODUCTION

The role of IoT technology in environmental monitoring has immensely revolutionized the way monitoring is done since it offers a rare opportunity of receiving and analysing data in real-time. One of the most significant transformations is observed in the sphere of forestry when IoT sensors are used for controlling vital indicators, including CO2 concentration and the level of moisture in the soil using ISMON and Sensecap data (Kashyap et al., 2021). These parameters are critical for demographic data, the conditions of an ecosystem, and the measures of a climate change effect. However, several prospects which apply to the use of IoT technology in this domain still exist, for instance; data quality, sensor calibration, data integration problems among others. This project seeks to achieve these challenges by providing efficient algorithms in data processing and machine learning for sensor calibration to improve the precision in the monitoring systems for the environment.

Figure 1: IoT Applications for CO2 in forest and Soil Moisture Measurement

1. 1 Background

The fact is that forests are essential for the world regime as they are capable of removing CO2 from the atmosphere and depositing it as biomass and humus. It plays a significant role in the reduction of climate change because it assists in reducing CO2 emission from activities carried out by humans. Continuous measurement of CO2 within the forests helps determine the processes of carbon storage and forests’ contribution to the global carbon balance (Salam  et al., 2020). soil moisture affects the forests and their health by controlling plant growth, nutrient turnover, and equilibrium in the ecosystem. Sensors are accurate in measuring these parameters, however, the ability to monitor these parameters depends on the IoT sensors used which need to function under various conditions that can be sometimes unfriendly.

The use of IoT sensors in forest area has specific issues to consider. Environmental conditions such as temperature, humidity, movement and vibrations can disturb the sensors’ stability and thus create problems in the data quality including low sensitivity, presence of many outliners, and issues of calibration. In addition, the obtained data from various sensors can be disrupted by the problems of timestamping and sensor drift (Marcu et al., 2019). These challenges call for increased use of sophisticated methods of data processing and calibration in the collected data to reduce on the credibility of the results obtained.

1. 2 Problem Statement

It is implicit as to what the main concern of this work is – data quality of IoT sensors monitoring CO2 levels and soil moisture in forests. Such factors as environmental conditions and operational conditions cause some defects in these sensors, such as outliers (or random errors), drift, and errors in measurements (Patrizi et al., 2022). These problems reduce the application of the data for decision making and other analysis. Thus, it become imperative to establish proper techniques of identifying these staleness and means of rectifying them with accurate data for environmental uses and to be processed for specific functions.

1. 3 Aim

The actual objective of this project is to develop, employ and promote methods that will improve the precision of the IoT sensors employed in forest CO2 and soil moisture measurement (Keswani et al., 2019). This includes methods for the identification of anomalous data observations, verifying the sensors’ calibration, and the data fusion, using statistical and machine learning techniques to enhance and validate the collected data.

1.4 Objectives

To achieve the aim of this project, the following objectives have been identified:

1. To design an algorithms for detecting outliers in IoT sensor data for forest CO2 and soil moisture monitoring (Drakulić et al., 2020).
2. To develop calibration techniques that can improve the accuracy of CO2 and soil moisture sensors under varying environmental conditions.
3. To integrate data from multiple IoT sensors, addressing issues such as timestamp discrepancies and sensor drift.
4. To evaluate the performance of the proposed methods using real-world sensor data from forest monitoring projects.
5. To provide recommendations for the deployment and maintenance of IoT sensors in forest environments based on the findings of the project.

1.5 Research Questions

The following research questions guide the project:

1. How can outliers in IoT sensor data for forest CO2 and soil moisture monitoring be effectively detected and corrected?
2. What calibration techniques can be employed to enhance the accuracy of IoT sensors under diverse environmental conditions?
3. How can data from multiple IoT sensors be integrated to provide a coherent and reliable dataset for forest monitoring?
4. What are the best practices for deploying and maintaining IoT sensors in forest environments to ensure data quality and reliability?

1.6 Significance of the Study

There are several research implications inherent in this study. First of all, it is to emphasize the lack of data on the monitoring of environmental changes in forests that are considered as a significant part of the world functioning as the CO2 sink. Due to the reliability increase of the IoT sensors, the present investigation contributes to the formulation of optimal decisions regarding forest management, nature preservation, and combatting of climate change. Therefore, accurate data on CO2 levels and other parameters such as moisture in the soil can assist in determining the state of health of the forest and other parameters in probably changing ecology and enforcing measures for sustainable management.

Secondly, further to knowledge in the Internet of Things technology specifically, the study presents superior methods of data processing and calibration that can suit a range of environmental monitoring projects (Minget al., 2019). The methodologies introduced in this project can be transferred to another application field, as IoT technology can be integrated in other applied domains including; agriculture, monitoring of air quality in cities, water resource management, and many more.

1. 7 Importance of the Study

The significance of this research study is specifying the objectives that aim at improving IoT technology for environment tracking. The accurate and genuine information that is provided by the IoT sensors is critical in capturing the dynamics of ecosystem and evaluate the effects of climate change while at the same time creating policies to advance forest management. Through tackling the issues related to the variability of the sensors’ precision and the data quality, the study will enhance the dependability of the environmental monitoring systems.

In addition, the paper discusses the importance of the complex data analysis and application of machine learning to enhance IoT sensors’ accuracy (Madushanki  et al., 2019). This, in turn, helps both to develop the field of environmental monitoring and expand the knowledge base of data science showing its usage in practice.

1.8 Need for the Study

Furthermore, there is a need for this study because the use of IoT sensors especially in monitoring the environment has continued to increase and with it the need to address issues that give rise to quality data is paramount. Some of the practical consideration during the usage of IoT sensors especially in forest environment include the following considerations. Procedures regarding calibration of the sensors together with the methods used in data processing are critical in the sense that the collected data have to be meaningful and reliable in support of the scientific and managerial goals (Maraveas et al., 2022). Besides, the rising effects of climate change and the essential part of forests in addressing these consequences show the need for proper monitoring of the environments. For this reason, accurate data on the levels of CO2 and moisture in the soil to assess the state of the forest, to predict changes in the structure of the biocoenosis and fulfill the management of the relevant activities. This study shall seek to fill these gaps by designing highly reliable approaches to enhancing the precision of IoT sensors utilised in forestry surveillance.

CHAPTER 2

LITERATURE REVIEW

2.1 Introduction

The world has seen the power of IoT through connected things by implementing real-time monitoring in several fields. In environmental control, IoT applications are particularly relevant for monitoring and controlling important parameters including CO2 levels, moisture etc. Measuring CO2 is crucial for combating climate change and enhancing industrial productivity on the other hand measuring soil moisture is critical for agriculture, particularly in the management of water. The utilization of IoT-based solutions for these parameters has additional options for real-time and remote monitoring, which will be beneficial for improving the environmental conditions and agricultural productivity along with the right decisions.

2. 2 Internet of Things in Measurements of CO2

Marques et al., (2019) that the current progress in several fields like Ambient Assisted Living and the Internet of Things (IoT) enables the creation of smart objects with rather high levels of sensing and connectivity. This paper introduces the iAirCO2 system which is an IoT architecture solution for the real-time monitoring of CO2. The iAirCO2 is formed by a hardware prototype for the measurement of the external context and a WEBSITE and smartphone application for consultation of data. In future, it is envisaged that these data may be retrieved by doctors to enhance diagnostic work. Compared to the other options offered, the iAirCO2 utilizes open source technology, resulting in total Wi-Fi systems with the following benefits: modularity, scalability, relatively inexpensive and easy installation. The obtained results show that it is possible to provide a feasible IAQ assessment using the presented system, which in turn enables further technical actions in favour of the improved dwelling conditions.

Bilotta et al.,(2022) researched that the Smoke from burning fossil fuels creates a proportionate input in changes to the atmospheric conditions and climate interferences. In cities, the contribution of traffic to CO2 is highly relevant and the total CO2 assessment can be made by (i) applying the fuel conversion to energy involving some factors and aspects of efficiency of engines and (ii) by considering the amount of weight to be transported, distance, time and emissions factor of the individual vehicle. Those approaches are inadequate for assessing the impacts of vehicles on CO2 in cities given that every vehicle emits CO2 relative to its efficiency, producer, fuel, weight, driver conduct, roads, seasons, etc Nowadays through modern information technologies, information on realistic traffic can be recorded to obtain useful information that can help in tracking changes in carbon footprint. The research conducted in this paper sought to identify the cause of CO2 emissions in various traffic patterns. In particular, the present paper suggests the model and the approach of CO2 emissions calculation based on traffic flow parameters considering uncongested and congestion conditions. They compound differently the total quantity of CO2 emitted into the atmosphere and hence give a different emissions factor.

Zhu et al., (2022) reviewed that the offers an effective solution regarding the goals of the environment monitoring of carbon dioxide combined with the IoT and cloud technologies. The named techniques will ensure an extremely high level of data openness and real-time visualizations that are critical for the efficient analysis of Smart Homes and actualization of efficient counter-measures. It was planned to have a monitoring architecture of the system which involves MQ135 carbon dioxide sensor as the core for generation and quantification of the amount of emitted carbon dioxide concentration, ESP8266 Wi-Fi module as the communication platform, Firebase Cloud Storage Service for accumulation and storage of the data and finally, Carbon Insight, an Android mobile application for visualization of the content of carbon dioxide concentration.

Venus et al., (2019) explored that the factors that produce healthy quality of accommodation were clean and healthy quality of air; unhealthy and polluted quality of air affected everyone who inhaled it in the population. In this research, the author has employed IoT technology to monitor the status of the quality of air temperature, and humidity of Carbon monoxide and Carbon dioxide. It uses ATmega328P-AU as the controller, DHT22 for air temperature and humidity, MQ-7 for CO gas, MQ-135 for CO2 ga, LPWAN LoRa as the data communication and data transfer medium and Antares as its cloud service through which all the data collected for display on an Android phone is stored.

2. 3 Smart Soil Moisture Measurement using IoT

Irawan et al., 2021 examined that irrigation in agriculture and the effective yield of crops depends on the soil moisture that is in direct proportion to the irrigation done. Therefore, measurement of soil moisture content is critical and this can be done using a soil moisture sensor. In this study, all the previous research carried out in the past 2-3 decades on soil moisture sensors have been reviewed and the principles of the most used types of soil moisture sensors and their various uses explained. Moreover, a comparison analysis of some assessment methods used was made on their merits, demerits, and factors that have a bearing on them. These improvements were presented by several scholars who have outlined the major use and performance criteria of soil moisture sensors, thus laying the groundwork for future research. 

Naresh et al., (2019) researched that in the early days, farmers tended to calculate the readiness of the ground and evoked suspicions that led to which kind of crop. They did not predict the condition of humidity, level of water and especially climate conditions which affected farmers increasingly The Internet of Things (IoT) is revolutionizing agribusiness by supporting agriculturists through a wide variety of approaches such as accuracy and precision farming to solve problems in the field. Assembly information on situations such as climate, dampness, temperature and fertility of the soil is made easier by IOT modernization Crop web-based assessment enables the identification of weeds, water level, insects, intrusion of animals into the field, vegetation profile, and horticulture. IOT use farmers to get related with their residents from wherever and at whatever point. Silage monitoring structures are used for observing the homestead conditions and lesser sized controllers are used for controlling and automating the home forms. For the remote observation of the conditions as picture and video, remote cameras have been employed. IOT development can reduce costs as well as update the productivity of standard development.

Doshi et al., (2019) Introducing technologies for sensing is one of the major processes to make crop production more sustainable through precision agriculture. Common sensing techniques suitable for the measurement of soil moisture and nutrient contents together with the sensing instruments which were recently benefited from the developments stated in the literature using those techniques are briefly discussed in this paper. Part one, ‘Soil Moisture: Measurement and Analysis’, has been further subdivided into four chapters explaining measurements of soil moisture parameters and laboratory tests, the in-situ tests, remote sensing and proximal sensing. All the above-mentioned technologies: their application, advantages and limitations are described.

Madushanki et al., (2019) IOT is the current and future of every field and touches the life of everyone by making almost everything smart. Several various devices form a self-organising network. Smart Farming with the use of IoT and the new developments of it by day are revolutionizing the traditional process of agriculture not only making it efficient but also reducing the cost of the farmers and minimizing crop losses. It is to provide an objective for which the technology under development may be employed to deliver the messages through one or another communication means to reach the farmers. The product will enable the farmer to get real-time details (Temperature, humidity, soil moisture, UV index, IR) of the farmland so that appropriate measures can be taken to carry out smart farming and hence enhance the production of crops and at the same reduce the use of resources like water and fertilizers.

Doshi et al., (2019) reviewed that there is a necessity to improve the productivity rates and effectiveness of the agricultural and related farming cycles and systems with the use of for instance IoT devices. First of all, the IoT can bring improvement to the agricultural and farming industry’s processes by decreasing the level of involvement of people due to the emphasis on the use of automation. The goal of this work is to provide a rather new overview of IoT applications that have already begun to emerge in the agriculture and farming industries, to explain the sensor data collections, the technologies and the sub-sectors such as water management and crop management that are related to this area. The data is taken from 60 reviewed scientific journal articles published in the year 2016-2018 covering the sub-vertical of IoT of the collection of sensor data for measurements for right decisions. As indicated in all the reported studies, this research finds that water management is the leading sub-vertical with a contribution of 28. 08%, the crop protection sub-vertical occupied 14. 60% contribution,

2.4 Integration of CO2 and Soil Moisture Sensors in IoT Systems

Patrizi et al., (2022) explored that precision farming technologies are, therefore, a range of modern tools and approaches used to improve operations on the plantation. Some of the advanced systems that have been adopted from an Agriculture 4.0 view are smart meter devices, IoT technologies, WSNs and so on. 0 points of view. Scholars of the recent literature have focused on the subject of AI and DL for assisting farmers and enhancing the quality of the soil. To this end, this article describes the design of a WSN powered by low-cost, low-power PV-supplied sensor nodes with the capability of monitoring information about the environmental condition and the soil content. Jensen, Costet, and Camillo concluded that the most challenging for implementing sensors is the soil moisture sensors due to the number of problems regarding their cost, installation, reliability, and calibration.

Hassan et al., (2023) in this article, give an account of the design and operating mechanism of an internet-of-things for the detection of soil CO2 levels. The global concentration of CO2 in the atmosphere has been on the rise and therefore the exact measurement of the major carbon reservoirs like soil for purposes of land management and policy making. Therefore, a set of IoT-connected probes for CO2 measurement in soils was designed in batches. These sensors were developed to measure the distribution of CO2 concentration across a site and report to a central gateway through LoRa. CO2 concentration and other environmental parameters such as temperature, humidity and concentrations of volatile organic compounds were recorded locally from the sensors and the data was transmitted to the user via a mobile (GSM) connection to a hosted website. After three deployments of the field in summer and autumn, a tendency for soil CO2 concentration to increase with depth and its variation with the diurnal cycle within woodland systems was apparent. To decide this, we assessed that the unit could store data for an optimum of 14 days in the continuous flow of data.

Bouali et al., (2021) derived that the SWM involves the efficient utilization and management of water-table or groundwater through the monitoring of water use through an IoT technology that is Cloud-based in nature. Renewable-energy integration supports energy-efficient agriculture as the use of fossil fuels in pumping water tables and more. The objectives of using Smart Irrigation will be; to ensure the crops receive adequate water nutrients to enhance the quality as well as the quantity while at the same time being conscious of the impact that this exercise has on the quality of the soil and water-table ecosystems. This research has been implemented and evaluated experimentally using a real Smart Farm system. Data obtained in this study pointed to the fact that the use of the proposed SA systemic approach minimises water consumption (using the adopted traditional irrigation system) by up to 71%. 8%. Last but not least, our solution has been kept open source so that it can be easily integrated and extended by other researchers to further enhance the cause of setting up a specific Cloud-based platform for water-table usage in the arid and sub-Saharan region.

Ramson et al., (Ramson et al., 2021)The regular diagnosis of soil health involves complex field surveys and laboratory tests. While this may provide correct results, it is quite expensive and time-consuming and not ideal for monitoring the changes in the state of the soil frequently. With soil sensors and wireless technologies rapidly becoming miniaturized and more accurate, the current off-line measurement and physical sampling are expected to be substituted by continuous monitoring in the field. This article presents the design, implementation and evaluation of an IoT-based smart system for monitoring soil health. The terminal nodes of the envisaged system are called Soil Health Monitoring Units and are solar-operated to be deployed in a field for a prolonged time. Every SHMU measures provides and sends data about soil temperature, moisture, electrical conductivity, carbon dioxide (CO 2 ) and geolocation over the air using LoRaWAN radio technology of a long-range wide area network. Such information is transmitted by a LoRaWAN gateway for its further uploading to server storage and analysis. A Web-based solution designed to show the data acquired is a feature offered to users.

Pithadiya et al., (2023) examined the developed to measure and manage various factors in a greenhouse through IoT, this system was created with a remote in mind. The multiple-point system requires sensing of three essentially important parameters; moisture content in the soil, the ambient temperature and humidity (ten moisture sensors, four Humidity as well as Temperature sensors for internal environmental Triggers the Exhaust fan and Water pump for maintaining the ambient condition inside according). This system was tested in a cucumber greenhouse environment and revealed that it was tracking the number of parameters on a real-time basis and recording them in the database. It only managed to moderate the parameters to levels that were within the prescribed norms. This system costs low is easy to use for the farmer and the farmer experiences an improvement in agriculture yield.

2.5 Research Gap

However, there are still some research limitations that bar the full effectiveness of IoT technologies for CO2 and soil moisture measurement systems; these include: Include problems with the accuracy and variability of low-cost sensor data that we can barely attain an acceptable level of precision in a fluctuating environment. Furthermore, combining data from various sensors into coherent understandable reports is still a challenge especially when dealing with large-scale systems and data integration problems. We also require better algorithms for processing big data and analytical models that will capture the complexity of the data from these IoT systems. Moreover, the issues of lifetime and management of IoT devices in remote and stringent environments are also limited. These challenging areas demand new ideas regarding sensors, data fusion, and the durability of the IoT system to improve the applicability and usability of IoT devices for CO2 and SM detection.

CHAPTER 3

RESEARCH METHODOLOGY

3.1 Introduction

The chapter on the research method explains the systematic approach that has been used to meet the research objectives and questions. This chapter gives an exposure of the total research proposal including research methodology, data collection techniques, and data analysis tools & instruments through which relevant information is fetched from the data. It is an elaboration of the approach used in the research about the justification of the selected methodological approaches and their application. This chapter therefore presents a framework of the research design, data sources, data preprocessing, analytical methods as well as an ethical consideration in handling the study.

3.2 Research Design

The method used in the conduct of this research study combines both quantitative and qualitative research approaches to attain a comprehensive analysis of the research topic. The quantitative strategy works therefore with numbers to analyze factors, or coefficients, and significance or probability, offering a general view of the relative or absolute relationships with the data collected. It is especially useful when dealing with large amounts of data and making probable inferences (Bersani et al., 2022). Besides this, a qualitative method is used to give the findings a broader perspective and more context. This dual approach guarantees a rich analysis of data, which is beneficial in minder interpreting the results attained in the Research Study.

The research strategy that has been used in this study comprises both descriptive and correlational research designs. Descriptive research is conducted to give an account of the variables being studied and to offer information on their features and dispersion. On the other hand, correlational research investigates the relationships of several variables to establish important patterns and connections. This strategy is quite useful in determining how levels of CO2 are linked to volumetric water content (VWC) and many other factors. Therefore using the aforementioned research approaches, the study’s purpose is to catalog and measure the state of affairs of the relationships between the focal variables and lay a base for subsequent research.

Some of the goals of the study are as follows: investigate temporal patterns of CO2 concentration, establish the link between CO2 concentration and VWC, investigate the effect of temperature and other factors on CO2 concentration, and construct the forecast model based on historical data. These are the overall objectives that are used to direct the research and choose the methods and methods for the research. By achieving these objectives, the current study seeks to offer important findings on the behaviour of CO2 levels and its determinants.

3.3 Data Collection

The data collection process for this study involves the use of two key datasets: CO2 levels and volumetric water content datasets were analyzed as the independent datasets, namely the CO2 levels dataset and the VWC dataset. The sources called CO2 levels consist of the readings of the CO2 concentration discovered by managing environmental sensors, with features such as Timestamp, DeviceID, and CO2 flogged in parts per million. The variables of the VWC dataset are composed of Timestamp, device_ID, VWC, Temperature, and Battery, and the dataset includes data that characterizes volumetric water content and related metrics. The two databases are from measurement systems designed to sample environmental conditions thereby providing realistic data to the surveyer and a broad representation of the environmental conditions being investigated.

Techniques used included placing environmental sensors in several sites to record values for CO2 concentrations, VWC, temperature and battery status. They made the sensors to be standard to allow a degree of precision in the monitoring process. It was collected in discrete time steps and written to raw CSV files which are plain text files containing tabular data. As such, it promotes the structured and standard process of collecting data which is fundamental to analysis.

To ensure that the collected data could be retrieved and analyzed is relative ease, practices of data storage and data management were put in place (Kashyap and Kumar, 2021). All the data files collected and produced were named and arranged logically according to their content and their source. Data backup was done daily so that in case of loss of data, the original contents could be retrieved for use in future. They add value to the management and integrity of the datasets as is seen below.

3.4 Data Analysis

The data pre-processing stage is vital in the data analysis since it prepares the data for the process. It encompasses several mission-critical activities, some of which are data cleaning, transformation, and integration. Data pre-processing consists of data cleaning that deals with missing values inconsistent data, and erroneous data. To deal with missing observations, procedures like forward filling were used, so that consecutiveness of the time series data will not be interrupted. Data cleaning also involved converting the time stamps to a uniform datetime format and checking that all the variables were in the right format. Data integration included joining of the data from different sources according to the similarity criteria that included timestamps and the device IDs and gave the consolidated view of the environmental conditions.

A preliminary analysis of the data was done to get first insights into the data and for this purpose, Exploratory Data Analysis (EDA) was performed to look for characteristics such as Structures, tendencies or outliers, in the data. Tabular data such as measures of central tendency and dispersion were computed regarding CO2 concentrations, VWC, and other parameters. Descriptive statistics describe the data set and are used to detect outliers and other peculiarities of the data (Keswani et al., 2019). Visualization belonged to EDA as many plots and charts were generated to visualize data and dependencies. Among those were line plots, histograms, box plots and scatter plots which helped to recognize the existing trends and patterns.

Data analysis that involved time series was used because it involves the analysis of data obtained at different time intervals to determine trends, cycles and seasonal changes. Some of the methods employed under time series analysis include; trend analysis, decomposition of trend and seasonality, and forecasting models. Trending included the plotting of data against time and analysis of movement in the trend line to determine long-term movements. In seasonal decomposition time series data was decomposed into an underlying trend, seasonal, and residual components; thus revealing the impact of periodic effects and noise. For long-term forecasting of the future levels of CO2, the following analytical tools were employed: Predictive modelling of the future levels of CO2; using the methods of ARIMA (AutoRegressive Integrated Moving Average) and SARIMA (Seasonal ARIMA). These models assist in the identification of the patterns and, therefore, assist in extrapolation to future patterns that would be expected.

Descriptive statistics were used in the evaluation of data associativity and co-efficiency factors. The coefficient of correlation analysis included the use of either Pearson’s or Spearman’s correlation coefficient to estimate the type and magnitude of associations between CO2 and other factors. Linear regression as well as multiple regression models were employed to determine the effect of parameters like VWC and temperature on CO2 concentration. Such analyses give information about the ability of these variables to predict CO2 changes and contribute to the knowledge about the behaviour of this factor.

There was inter alia the use of matrices to improve the analysis as well as the use of detailed mathematical computations to get enhanced results. The latter played a role in the Vector Autoregression (VAR) method that allowed analysis relationships of time series variables, related to the way the change of the particular variable influences the change of other variables in future. In analysing the time series, techniques like Exponential Smoothing (Holt-Winters Smoothing) were used to filter out unwanted noise and look for trends. Autoregressive distributed lag analysis was used to consider the directionality and strength of the effects of multiple time series variables on one another.

3.5 Model Evaluation

It is worth mentioning the evaluation of the models is an important process of ascertaining the accuracy and validity of the analytical models applied in the paper. When selecting the model, some criteria are used to measure the performance of the models These are the fit criteria and accuracy of the models (Madushanki et al., 2019). To justify the choice of models in predicting the past data, AIC (Akaike Information Criterion) and BIC (Bayesian Information Criterion) were used. Regarding performance, Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) benchmarks were used to measure the predictiveness of the models generated for forecasting.

Model validation implies verification of whether the selected models are well suited for new data. Cross-validation applies the simple train-test method of splitting a set of data into the training database and the testing database to test the model performance. This process aids in validating the model as to the appropriateness, strength and reliability that is required. Using the residuals of the model, one can always check for the existence of systematic patterns or biases that may be affecting the model fit or the assumptions that have been made.

Use of the sensitivity analysis to show the impact on the results of modification of model parameters. The activities included in parameter tuning were the fine-tuning of the parameters that were used in the model while in scenario analysis the model simulated different scenarios to quantify how different inputs affect the model. Sensitivity analysis gives actual information regarding which are the most sensitive parameters and how they contribute to change model results.

3.6 Ethical Considerations

Issues of an ethical nature are very essential in doing research, especially in dealing with data and methods to take whatever measures to ensure that the data collected is secure. Privacy was another issue of importance and steps were taken to ensure that data privacy was achieved individual’s privacy. Methods such as ‘blanking out’ or ‘masking’ were applied to reduce or eliminate identifiable data so that the information cannot be directly associated with individuals or places concerned. Anti-data loss management procedures such as encrypting data and limiting the access of data to unauthorized persons were also put in place to prevent leakage of data.

Ethical use of data entails both informed and express permission to be given. It also entails that all the steps as well as findings of a given study are well documented in the public domain. It is crucial to have the necessary permits and consents regarding the collection and usage of data, which will help with following the code of ethics and laws (Maraveas et al., 2022). Thus, is important that the collection of the data is done complies with the set ethical standards so that the research does not suffer from biased data.

Policy all of the implications of the findings of the study should be given attention. It is crucial always to evaluate how the findings might inform the stakeholders such as policymakers and the general public or environmental organisations to ensure that reporting is done responsibly. Appreciation of the likelihood of the results contributes to decision-making and recommendations regarding the observations obtained in the study. That is how it is possible to define some insights for further investigation based on the existing findings and to suggest ways to expand the investigation and if necessary fill the gaps in the study.

3.7 Limitations of the Study

In any research study, some limitations are inherent in the study that can affect the generality and usefulness of the findings. These considerations are important to provide a comprehensive view of the outcomes of the research as well as to orient further investigations. The following are the potential sources of bias, the gaps in data quality, the modelling methodology and the externality of the study.

1. Data Quality and Completeness:

The first and arguably the most significant limitation is in respect of the quality and completeness of the data used for the research. When it comes to the application of the model to a new study, various challenges impact the strength of the conclusions Such a situation may be explained by the presence of missing values, sensor errors, and various data irregularities. Whereas in the case of forward filling, there was an effort to deal with the problem of missing observations, then imputation of missing values can result in biases or other errors in the analysis. Moreover, due to sensors and, respectively, the inaccuracies in calibration, the readings of measurements can also differ from each other.

2. Temporal and Spatial Coverage:

Another weakness of the study is the fact that the research covers a rather long period and a vast area. The information was collected at certain times and in certain conditions, thus not giving a complete view of the environment. For example, changes in CO2 and VWC might be affected by local climate conditions, land use, or vegetation, which might have been different in various sites and thus were not described to cover all the data. That is, the limited physical scope may also inhibit the generalization of the study outcomes in other areas or systems with different environmental conditions.

3. Methodological Constraints:

The current study is also affected by the constraints of methodological measures. The selection of the analytical tools and the modelling itself is derived from the best practices, hence, it may not capture all issues in the data. For instance, while the forecasts derived from ARIMA and SARIMA models are efficient, they might not capture all of the non-linearities that occur or all of the changes in the CO2 levels (Marcu et al., 2019). Likewise, correlation analysis fails to capture the actual matrix among variables because, for instance, it adopts linear computations rather than nonlinear computations for variables that depict absolute/non-absolute correlations.

4. Data Integration Challenges:

Combining information that has been gathered from various sources is not without its difficulties, the effects of which may be felt in the analysis that is conducted. Variations in attribute formats, time stamps and units of measurement to name but a few can pose a challenge in the merger process and might result in the formation of an erroneous connected data set. Nevertheless, variations in the formats of the data retrieved from different sources may affect the validity of the combined data and the results of the evaluation studies.

5. Model Assumptions and Limitations:

The models employed in this paper including the ARIMA, SARIMA as well as the VAR rely on some assumptions about the data for instance data stationarity and linearity. Inaccuracies of this assumption can impact the accuracy of the model and hence its predictions. For instance, the ARIMA model assumes that the time series is stationary, and any non-stationary condition may cause a wrong forecast. Likewise, the accuracy of the SARIMA model depends on the right specification of seasonal if there are any, but this can be imprecise.

6. External Factors and Uncertainty:

Lack of control over factors outside the social research study, for example, alteration in environmental plans, technological developments or natural calamities, may create a great number of variables in the results of the study (Ming et al., 2019). For instance, a shift in the techniques used in sensors or data capturing methods may affect the history and present data. Further, the interannual variability of CO2 observed in Figures 2 and 3 might be affected by other factors that alter the environmental conditions and therefore affect CO2 and VWC in unexpected ways.

7. Generalizability of Findings:

It should also be borne in mind that, these conclusions may have low external validity because the data for analysis was obtained in a particular setting and under certain circumstances. Although the approach of the use of statistical models gives information on how the levels of CO2 relate to VWC in the specified area of study, there might be limitations in its generalization of others over other areas in dissimilar environments. However, more research has to be done to replicate the study and check the generalizations in different settings.

8. Limitations in Predictive Accuracy

The predictive models built in this study while producing useful predictions inherently suffer from the errors of prediction in terms of accuracy and precision. The problem is that forecasting models are far from perfect, and they are capable of providing rather approximate estimations for future changes. The performance of a model for predicting is a function of c parameters, data, and the occurrence of other conditions. Such forecasts, despite being very useful, should be taken with a pinch of salt due to the likely margin of errors of deviations from data.

3.8 Summary

In summary, Chapter 3 presents a descriptive analysis of the research methodology used in the present study. This paper covers the research methodology, data acquisition approach, data analysis strategy, and model assessment parameters. The research procedures adopted in the study incorporate both qualitative and quantitative research methods to enrich the analysis of the findings on CO2 and other factors (Patrizi et al., 2022). The overuse of analytical techniques and models means that the results of the study are more reliable. Some of these include: Ethical issues are considered to make sure that the research is conducted in the right manner. In the last chapter, the author gives an overview of the contributions made by the research method and presents a research agenda and recommendations for the empirical and practical use of the method. This chapter offers a rational and comprehensive method of dealing with data and thereby helps in setting the pace in the research endeavours and their results.

CHAPTER 4

DATA ANALYSIS AND VISUALIZATION

4.1 Data Loading and Preprocessing

4.1.1 Introduction to Data Sources

The analysis centres around two primary datasets: Basic measurements that were recorded include; CO2 levels and Volumetric Water Content (VWC). The CO2 dataset comprises the following fields: Timestamp, DeviceID, and co2 in which Timestamp, DeviceID, and co2 signify the date and time of measurement, identification number of the measuring device, and the concentration of CO2 in parts per million (ppm) respectively. VWC dataset includes Timestamp, Device_ID, VWC, Temperature, and Battery which are the time of sampling, unique identification number of the device, percentage volumetric water content, temperature in degrees Celsius and battery level of the sensor respectively.

These datasets are in the CSV file formats – the most basic file type for tabular data owing to its compatibility with a wide range of data processing tools (Bersani et al., 2022). The CO2 levels and the VWC measurements were obtained from environmental sensors which were placed in different locations to capture the environmental variations over time.

Table 1: Data Loading and Preprocessing

4.1.2 Data Merging and Integration

Coaxing data from one source with another usually means putting data that was stored in different files into one unified and easily manageable data frame. In this case, it was necessary to correlate the obtained values of the CO2 and VWC to make the merger By the time of obtaining the data set, their timestamps were synchronized. Firstly, one DataFrames for each of the datasets were generated in the usage of the pandas utility. By utilizing the merge() function, the internal datasets of the two files were merged based on the Timestamp field and DeviceID/Device_ID respectively. This way of data processing prevents the situation when data from different sources are collected at different times or for different devices, though, they have to be considered at the same time and for the same device.

Dealing with different files and joining big data files needed utmost care. The details of data preprocessing included the conversion of different formats of timestamps into a common format and the alignment of device IDs. Appending the two tables combines the CO2 level data with VWC readings to improve combined interpretations of how CO2 concentrations may change about VWC and other factors such as temperature.

4.1.3 Handling Missing Values

This is always a challenge especially when data is missing because lack of it weakens the quality of the analysis that one intends to make (Kashyap and Kumar, 2021). As for data gaps in the datasets given, forward filling was applied, it occurs when a dataset has gaps, each of which is replaced by the last accurate value. This technique is based on the presumption that missing values are similar to the values that were recorded before; this is a presumption that holds particularly where data is periodical.

This minimized the effect of missing data in the analysis but one has to be careful of the biases that one is introducing. For instance, forward filling might make the device assume certain environmental conditions that prevailed during a certain period when the device was not recording data at all in the shortest time possible. Understandably, one must perform exploratory analysis to determine the degree and distribution of the missing data to assess its possible effects on the outcome.

4.1.4 Data Consistency and Conversion

Interformat consistency is important to obtain the right results. Casting the timestamps into datetime format helps in the correct computation of any time-related operations like ordering and filtering. This conversion helps in much easier calculation and analysis of time-series data, which makes trend analysis and temporal comparisons more precise.

To ensure data consistency, parsing errors were addressed; if there were discrepancies, they were also corrected. For example, overlapping or dissimilar device identifiers, or incorrect temperature measurements were handled during data cleansing. The preprocessing of data involves cleaning and formatting, where all the data is made to maintain a standard format.

4.1.5 Summary of Preprocessed Data

The integrated and cleaned dataset shows all the environmental measures captured by the sensors in an elaborate preprocessed format. Some of these statistics include average CO2 levels, mid-point VWC percentages, temperature variations and so on. The cleaned dataset is now prepared to be analyzed in detail, and data gaps are also filled, and there is no inconsistency. Such preparation enables the firm to effectively monitor the levels of CO2 in correlation with the volumetric water content or other factors in the environment.

4.2 Data Visualization

4.2.1 CO2 Levels Over Time

The first visual includes on its axis the title “CO2 Levels Over Time” where a reader can observe different periods and their impact on CO2 levels. In this line plot, the CO2 data are shown for different timestamps where each device is depicted in a different colour. On the x-axis, we have time and on the y-axis, the levels of CO2. The use of different colours for each of the devices makes it easy to compare CO2 levels recorded by the different devices over time.

In this case, it is easier to observe temporal patterns and fluctuations of the CO2 concentration. For instance, it will be possible to observe that general rates of CO2 always rise or fall in a periodic cycle (Keswani et al., 2019). It also enables the identification of any abrupt changes or an increase or decrease in the levels of CO2 concentration, which may reflect problems arising with certain devices, or some interference with the measurements. Such patterns can play a crucial role in understanding seasonal or period factors and in making decisions based on temporal data patterns.

Figure 2: CO2 Levels Over Time

4.2.2 Distribution of CO2 Levels

Histogram – “Distribution of CO2 Levels” in conjunction with Kernel Density Estimate gives a visual outlook on the frequencies of the CO2 levels over the given set of data. The first one shows the number of occurrences of the CO2 measurements in the specified bins and the second one is a smoother curve of the histogram.

This visualization is relevant for visualizing the density of the general distribution of CO2 concerning geographical locations. What it does is provide information about how CO2 measurements are distributed across certain ranges and the various occurrences of certain CO2 levels. For instance, if the histogram plots have a hump in a given interval of CO2 levels, it means that these levels are often present in the dataset (Madushanki et al., 2019). On the other hand, if the distribution is bell-shaped with its curve curving downwards at both ends or if it has one or more humps then it might imply that the samples obtained of CO2 had different conditions or there could have been a problem with the measurements.

From the histogram and the KDE curve, the analyst can conclude the central and dispersive characteristics of the CO2 level, as well as its shape, to define changes in levels that are frequent in the set.

Figure 3 :Distribution of CO2 Levels

4.2.3 Comparison of CO2 Levels by Device

The “Boxplot of CO2 Levels by Device ID” presents the CO2 lengths in the form of a boxplot for each of the devised devices. The graph in panel A of Fig. 3 is a box-and-whisker plot, and each boxplot in the figure corresponds to a device with the horizontal position of a device indicating the median of the CO2 measurements for that device.

This analysis is useful for determining fluctuation in the level of CO2 in the different devices. From the box plot, one can infer information on the median level for each device and the dispersion or spread of the CO2 level as well as the possibility of the existence of outliers. This is because, LTE, IQRanges and LOutliers represent a measure of dispersion and variability of CO2 measurements that may differ because of variation in device performance or conditions that the device is used in.

From box plots, it is possible to determine which devices always provide higher or lower levels of CO2 and can easily detect any anomalous readings (Maraveas et al., 2022). The information produced by the devices can be used to validate the differences and to assess the influence of possible causes for the inconsistencies in the reading of the CO2 level.

Figure 4 : Comparison of CO2 Levels by Device

4.2.4 Correlation Heatmap

The “Correlation Heatmap” gives an overall view of how the variables – here, CO2 and height are related. The heatmap itself is created in such a way that the color codes point out correlation and positive or negative direction and the intensity of color defines the strength of correlation.

This visualization aids in accessing the relationship between CO2 levels with other parameters like Height. From the heatmap, one can see whether high or low CO2 is associated with certain values of height (Marcu et al., 2019). Cohesion may leave an impression that strong positive correlations mean that the variables share a potent relationship, while low to non-existent correlation means that the variables have no relation.

Interacting with the correlation heatmap helps understand relations between the variables and can be used for further analysis or some investigations regarding the causes for such relations.

Figure 5 : Correlation Heatmap

4.2.5 Height vs CO2 Levels

The “Height vs CO2 Levels” scatter plot aims at finding a correlation between height and CO2 levels. Each dot on the graph is an observation – height is plotted on the vertical ‘y’ axis and CO2 levels on the horizontal ‘x’ axis.

This chart helps analyze if there is evidence of rejection of the null hypothesis in the case of height and CO2 levels. For instance, if the emerging pattern of high heights, or low heights is associated with high or low levels of CO2, then it may be an indication of a relationship between these variables (Ming et al., 2019). The situation can also be seen from the scatter plot where we can find out if there are any outliers or unusual values that need to be looked at more closely.

It is possible to evaluate from the scatter plot if height has any observable effect on CO2 levels and if certain aspects have to be sought out for further investigations.

Figure 6 : Height vs CO2 Levels

4.3 Summary of Findings

The basic and extended visualizations give a good orientation in the dataset and allow to study of basic trends, distributions, and connections. The idea of using a line plot to exhibit temporal data such as changes in CO2 levels over certain periods is effective in identifying periodicities and aberrations as well as tendencies of CO2 Cleveland variables such as histogram and KDE curve is useful in extending the knowledge of tendencies in CO2 level. The boxplot shows the distribution of CO2 levels between devices as well as its within-device variation and the correlation heatmap as well as scatter plot reveal the sort of relationships between the other set of variables and CO2 levels.

4.3 Time Series Analysis

Time series analysis is another important statistical technique that is used in the analysis of data which is collected at different points of time. They are applied to incubate data gathered at predetermined intervals for the purpose of establishing trends and connections. This approach can be employed when deriving the future value of an item based on learned experience, it finds application in environmental science, economy and finance. This section also focuses at other techniques of time series including trend analysis, seasonal decomposition, ARIMA, SARIMA, exponential smoothening, simple moving averages, VAR, Prophet and the cross-correlation analysis. As in the previous method, it helps to better define temporal patterns of the dataset, as well as the dynamics of CO2 and its dependence on the observed factor, including VWC.

4.3.1 Introduction to Time Series Analysis

In time series analysis, the data collected is followed over a time cycle and attempts to search, find for some pattern. This analysis is necessary for dissecting temporal patterns of various aspects including environmental factors such as CO2 levels. The main goals of time series analysis are an estimation of the trends, seasonality, and cyclical behaviour with the view of forecasting.

Example of a simple time series graph of the perception of CO2 level over time. On the x-axis, time intervals are reflected whereas on the y-axis there is the CO2 concentration. In analyzing such plots we can see the general trends of CO2 levels and any abrupt changes or otherwise.

Through the use of different techniques in time series, they can find details that may indicate long-term trends, short-term variation, cyclical behaviours and the like (Patrizi et al., 2022). These ideas are valuable when it is necessary to forecast the CO2 levels and to comprehend the conditions that have an impact on it. For instance, observing an increase in the CO2 levels among the variables might notify more growing concerns about the environment while observing periodic cyclic patterns among the variables might suggest periodic influences such as seasonal changes in climate or human activities.

4.3.2 Trend Analysis

Trend analysis is done on data to determine long-term movements in the data with time. This tends to be about using graphic displays to identify trends that can be marked by gradual changes with time. It is useful for evaluating if the concentrations of CO2 are increasing, decreasing or stabilizing and, in addition, offers a perspective on environmental trends.

Figure 3. 3. 2. 1 summarizes the trend of CO2 amount for a few years. The plot also displays an increasing manner of the curve, meaning that the CO2 concentration increases with time as it is illustrated in the figure below. It is used to determine environmental policies, industrial activities or any other factor which has contributed to the rise of CO2 levels.

One of the most important ways to carry out data analysis is a trend analysis, which shows exceptions, jumps or declines from one period to another period. For instance, CO2 level increase could be related to specific activities for instance accidents or shifts in emission standards (Salam and Salam, 2020). In this respect, understanding these trends helps opponents, policy-makers, and researchers to act properly and prevent severe consequences of accompanying negative phenomena.

Figure 7 : Trend Analysis

4.3.3 Seasonal Decomposition

Seasonal decomposition involves breaking down a time series into its constituent components: These demographic trends include cookie trend, seasonal, and residual. This decomposition is useful in explaining how the seasonal factors affect the CO2 levels as well as in noticing any periodicity or fluke.

Figure 3.3.3.1 illustrates the degree of fluctuation of CO2 levels displaying the seasonal breakdown. The plot is divided into three parts: The three components are, namely, the trend component, the seasonal component, and the irregular component. The trend indicates the long-run changes in CO2 levels and seasonality shows irregular changes that happen in fixed periods for example yearly or quarterly. The remaining part is used to hold other remaining fluctuations that are not accounted for by the trend or seasonality.

Seasonal decomposition is particularly relevant when trying to model fluctuations of CO2 concentrations concerning certain seasons. For example, increases in CO2 concentrations are more likely to occur in the cold months because of the escalated use of heating systems, while low concentrations are likely to be recorded in the summer because of the reduced industrial activities (Soheli et al., 2022). From these components, inferences can be made on the components influencing CO2 concentration and hence proposed solutions for raising the accuracy of the forecast.

Figure 8 : Seasonal Decomposition

Figure 9 : SARIMAX Results

4.3.4 ARIMA Model

The ARIMA is one of the most common procedures used to forecast time series data. It combines autoregressive (AR) and moving average (MA) components and it has differencing to address non-stationary data. Since their derivation is based on converting a random process to white noise, then ARIMA models are particularly well suited for non-seasonal data although they can be used for many tasks.

Figure 3.3.4.1 shows the CO2 Levels and the ARIMA model which has been fitted to the data. These are the actual values obtained from the observed data and the values that were forecasted by the ARIMA model present in the plot, of which the fitted line is red (Bersani et al., 2022). The model performance checker uses the criterion including Akaike Information Criterion (AIC) and residual.

The process of identifying the structure of the time series is therefore facilitated and accurately modelled through the use of the ARIMA model with the view of making accurate forecasts and subsequently carrying out trend analysis. For example, if the ARIMA model forecasts an increase in CO2 levels, it might recommend sustained emissions or changes to the environment. Using residuals and AIC to assess the model is a good way of determining whether the model is suitable for forecasting, and recognizes appropriate patterns from the data.

Figure 10 : ARIMA Model Fitting

4.3.5 SARIMA Model

SARIMA (Seasonal ARIMA) is an enhancement of the basic ARIMA technique that deals with a time series that has well-pronounced seasonal variations. As a type of ARIMA method, SARIMA models are very suitable when there is seasonality in the responses and such cases include the annual fluctuations in CO2 levels caused by climatic conditions.

Figure 3. 3. 5. 1 shows the fit of SARIMA to the analysis of CO2 levels. The figure below illustrates the observed values as well as a line of the fitted values according to the SARIMA model where the green line depicts the SARIMA fitted values (Marques et al., 2019). The chart on the summarization of the SARIMA model offers more details about the parameters involving the seasonal factors and more so the overall fitness of the model.

While using the SARIMA model, seasonal factors are incorporated in the model unlike the simple model, hence it can model the seasonal changes in the CO2 levels well. For instance, the model could show that the levels of CO2 rise to their highest levels in certain seasons, for example, winter due to the use of heaters. SARIMA analysis assists in determining the degree of the seasonality of the data and in enhancing the degree of accuracy of the CO2 level prediction.

Figure 11 : SARIMA Model Fitting

4.3.6 Exponential Smoothing

Simple Exponential Smoothing and the Holt-Winters Exponential Smoothing Techniques are used in the smoothing technique where values aid in the short-term forecasting of the values of a time series. The above methods give more percentages to the recent observations so that they give a smoothed curve showing recent changes.

Figure 3. 3. 6. 1 reveals how exponential smoothening can be used for level readings on CO2. The orange line perhaps may be interpreted as the fitted values of the exponential smoothing model which shows the recent trends and the movements.

This technique is popular in analyzing recent trends and patterns in making forecasts, hence it is popular in exponential smoothing methods (Bilotta et al., 2023). For example, if recent values of CO2 depict a steep rise, then using exponential smoothing it is possible to give a forecast which depicts variation and can look into the future values. This technique assists in making short-run forecasts and explains the effects of contemporary trends on CO2 for the short term.

Figure 12 : Exponential Smoothing Model

4.3.7 Moving Average

This is a method where to level out time series data, an average value of a window size of say 30 days is taken. This technique is useful in partizing noise and making visible trends that were concealed before.

Figure 3.3 7.1 also shows how the moving average is applied to CO2 levels. The plot contains the actual values and the line of the moving average; the orange line is for the trend.

If moving average plots are made alongside the raw data, it will be possible to appreciate how smoothing distorts trends. These averages assist in spotting major shifts or trends that might be masked by noises of various short-run variations (Chen et al., 2022). For instance, the moving average can be used to identify a gradual increase in the levels of CO2, whereas the actual data will present significant fluctuations.

Figure 13 : Moving Average

4.3.8 Vector Autoregression (VAR)

The use of the VAR (Vector Autoregression) model is based on the analysis of time series variables between which the dynamic interactions are revealed. This model is appropriate to use when identifying the impact that the change of one variable has on other variables.

Figure 3. 3. 8. 1 illustrates an example of VAR with more than one element, concerning CO2 levels and VWC. These variables’ plot also reveals the interrelation between these variables and the impact of their lagged values.

Sensitivity analysis gives indications of the connection between the levels of CO2, VWC, as well as other variables (Vanus et al., 2019). For instance, the analysis might show that an increase in CO2 levels is equally accompanied by an increase in VWC as they all have an interrelated relationship. It assists in explaining the relationships within the various variables within the time series data and assists in decision-making.

Figure 14 : Vector Autoregression (VAR)

4.3.9 Prophet Model

The Prophet model, built on Facebook, is used for time series trend forecasting with a high-degree seasonal pattern and holiday variations. It gives point as well as interval forecasts and this enables one to predict future values and to evaluate the accuracy of forecasts.

Figure 3.3.9.1 demonstrates and describes the results of the Prophet model used for CO2 levels forecast. In the plot, the values are fitted, predicted, and the upper and lower bounds of uncertainty.

The Prophet model is useful for investigating the behaviour of such trends over the long term and predicting them based on such data. For instance, the company may use the Prophet model and it predicts rising in the next year CO2 level might suggest that there is a continuing change within the environment or in emissions (Irawan and Muzakir, 2022). The uncertainty intervals further allow for evaluation of the degree of accuracy of the forecasts and for anticipating fluctuations.

Figure 15 : Forecast with Prophet

4.3.10 Cross-Correlation Analysis

Cross-correlation analysis works based on two or more time series variables and also recognizes lagged effects and cross-variable impacts. This technique increases the possibility of understanding how one variable causes or happens simultaneously with another.

Figure 3.3.10.1 shows the coordination factor between CO2 and VWC. The plot shows the correlation at a range of lag zero through twenty when the pairs of time series are correlated.

Having identified the trends in the variation of the CO2 level, VWC and the other variables that were being measured, cross-correlation analysis enables us to understand how these variables vary concurrently (Naresh et al., 2020). For instance, the results that will be highlighted can show how fluctuations of VWC trigger a corresponding fluctuation in CO2 many weeks or months later. It assists in identifying the Volterra series dependency of one variable on other variables, thus enabling a more precise appreciation of the time series data based on the model.

Figure 16 : Cross-Correlation Function

4.4 Sensor Data for CO2 and VMC Sensors

CO2 and soil moisture data collected by the sensors along with its attributes like sensor number or position coordinates (X and Y) and the accuracy . It then scales up the accuracy values to a Mosteller format or a range that is good for visualization by multiplying each of them with 1000 then this creates a new column tagged ‘Size’ in the given data set for the better visualization of the data points in the plot. By applying Matplotlib, the code plots the figure of scatter type, where each point refers to a sensor and is in the field of the indicated X and Y coordinates. The size of each point is proportional to its normalized accuracy; the color corresponds to accuracy as indicated on the colorbar (Doshi et al., 2019). Moreover, text labels are placed on each point, which shows the corresponding IDs of the sensors, and to make it clearer, the axes are labeled, the plot has the title and a grid is enabled.

Figure 17 : Sensor Positioning and Accuracy

4.5 Outlier Detection

This section performs an elaborate outlier detection analysis of a given column denoted as ‘Moving_Time’ on a data frame if exists. It employs three distinct outlier detection methods: Yo and N quote four methods: Z-Score, Interquartile Range (IQR), and Isolation Forest. First of all, the Z-Score method derives the standard scores of the data points and marks the ones with the absolute Z’s greater than 3 as outliers. For the purpose of providing the results of the implementation of these outlier detection methods, the code produces three box plots side-by-side; one is for each of a different method where outliers are plotted in red-color while those that were not detected as an outlier are in blue color.

Figure 18 : Box plot with Isolation Forest outliers

4.6 Correlation

4.6.1 Data Period 1

This section offered a systematic analysis and diagram of the moving time data as related to different sensor devices in the period of time given. First, it defines a dataset variable including time stamping, device ID,’moving times’ and ‘positions’ all of which are later converted to a data frame. To simplify the temporal analysis, the timestamps are transformed into datetime format, and the year and month of these timestamps are then derived (Madushanki et al., 2019). It is then grouped by year, month, position, and device ID, where the moving time is summed for each category. This aggregated data is then regrouped into a pivot table, which allows for correlation analysis with the year, month, and position as rows and the device IDs as columns, and any missing cells filled with 0s. Once again, the code looks at the pivot table obtained and then calculates the correlation matrix for the same, which shows the quantitative affinity between the times of movement across the different devices.

Figure 19 : Position-wise Correlation Matrix of Moving Time (Period 1)

4.6.2 Data Period 2

This section computes and visualizes different moving time data of different sensor devices for a given second interval. Finally, the formation of a dataset is done involving timestamps, device IDs, moving times, as well as positions. Such an outcome of data preprocessing is then converted into the data frame, wherein timestamps form an essential component to temporal analysis; the stamps are converted into the datetime format. Therefore, the year and the month are taken from the time stamp so that the data can be suitably categorized based on time intervals. After that, the data is divided by year and month and position, and by device ID, and the moving times are summed for each combination of those values (Doshi et al., 2019). The aggregated data is returned to the pivot table, allowing for correlation analysis where year, month, and position were set as column indexes and device IDs as columns filling missing values with zeros. There is then the computation of the correlation matrix to determine the degree of correspondence of moving times between different devices.

Figure 20 : Position-wise Correlation Matrix of Moving Time (Period 2)

4.7 Comparison

4.7.1 Position Comparison of the 4 Sensors

This section enables the analysis of average moving time for different sensor devices to be done per position. First, set the data points to include the time stamp, device ID, moving time, and position, and then convert the data to data frame format. The data is then postprocessed by summarization of moving time by position ID and device ID by using group by unstack, which makes the organized comparison of moving time with the specific categories. For visualizing these aggregated results, a heatmap is created using Seaborn, in which the X axis relates to device IDs and the Y axis refers to positions (Patrizi et al., 2022). The average moving times are shown as colored bar plots, with collective labels to offer exact values and avoid confusion.

Figure 21 : Average Moving Time by Sensor and Position

4.7.2 Sensor Value Analysis

This section performs statistical analysis of moving time data for the sensor devices, wherein the moving time data is summarized in terms of mean, median, and standard deviation for each of the devices. First, the data is split among device IDs, and then the specific statistical functions are used on the ‘Moving_Time’ option with the help of the agg method, which in turn gives out a new data frame containing the statistics of the particular device. The columns of this summary DataFrame are then renamed for clarity, and these are the calculated statistics below (Hassan et al., 2024). A bar plot is used to show these results with the help of the Seaborn library, where the values of device ID are taken on the x axis and the statistical values are taken on the y axis. The bar plot is made more effective through the use of the melt function that transforms the data for ease of the plot and the side-by-side comparison of mean, median, and standard deviation of the various sensors.

Figure 22 : Sensor Value Analysis

4.7.3 RMSE Comparison between Sensors

This section is a section of a linear regression analysis to check correlation between moving time and timestamps of selected sensor devices. At first, a dataset containing the timestamps, the associated device IDs, and the corresponding moving times is created and then read into a Pandas frame. The time stamps are transformed to datetime format and are then transformed into date numbers where the date number is equal to the difference in seconds from the first date to the current date symbol (Bouali et al., 2021). This conversion helps in the process of modeling. For each particular sensor, the data is then subsetted, and a linear regression model is then fitted with the numeric timestamps as the predictor and the moving times as the response variable. Information regarding the moving times is predicted for the proposed local model using the fitted model; RMSE and RTH parameters are computed in order to assess the accuracy of the model.

Figure 23 : Sensor Moving Time Behavior with Linear Regression

4.7.4 RMSE Comparison Based on Position Changes

This section undertakes a comparison of observed RMSE of linear regression models fitted to moving time data obtained from a number of qualitative and quantitative sensors installed at different locations. First, the research performs the preparation and compilation of a set with time stamps, device identifiers, moving time intervals, and positions corresponding to the movements (Ramson et al., 2021). They are then converted to DateTime format and given numeric values based on the number of seconds between the first and each of the other times for regression analysis reasons. It then does the same thing for each of the unique positions and sensors within that position and fits a linear regression model using the numeric timestamps as the independent variable and the moving times as the dependent variable. When the model still fits, there are moving times predictions, so the RMSE is computed to assess the predictive prowess of the model.

Figure 24 : RMSE Comparison Based on Position Changes

4.8 Model Behavior

This section presents a linear regression analysis to establish the relationship that exists between the timestamps and moving time as captured by the various sensors. First, a set of data is selected, which includes time stamps as well as the device identifier and the moving periods. The format of the timestamps is changed to trirdatetime format, and for conducting regression analysis, a new variable is created, which is the numerical equivalent of the time sealed since the first API call was created (Pithadiya et al., 2021). The code sets each device id and then grabs a slice of the data set unique to each sensor for that specific id. For each sensor, the linear regression is built for the moving times, while the numeric timestamps are used as an independent variable. The means of moving times are forecasted using the fitted model, and the measures of accuracy—the root mean square error (RMSE) and the coefficient of determination (R2)—are also computed.

Figure 25 : Sensor Moving Time Behavior with Linear Regression

4.8.1 Analysis of sensecapVMC-096

This section commences a data analysis on a dataset that comprises VWC gathered from the SENSECAP sensor. First, the data set is read with the help of the Pandas package; the first lines of the data are shown to get an idea of what it will look like (Fenoglio et al., 2023). After that, column names are echoed to let one know the available variables in the dataset under consideration. To perform simplified statistical analysis, the describe() function is applied to this data, which computes the count, mean, standard deviation, or quartiles of numerical variables that would help the analyst get an initial impression of the dataset. To guarantee the quality of the quality of the data,, the code validates the number of missing entries in each column and gives a summary. In this step, it is important to determine the quality of the collected data before shifting to the next level of examination.

Figure 26 : Basic Statistical Analysis

Figure 27 : Distribution of Volumetric Water Content (VWC)

4.8.2 Linear Regression Analysis of VWC and Temperature

This section carries out linear regression analysis that tests VWC with or without temperature depending on the two parameters in the dataset. First, it tries to detect the existence of ‘VWC’ and ‘Temperature’, which are in the DataFrame. If both are present, the independent variable that is VWC, the dependent variable that is Temperature is defined. After that, the linear regression model is created with the help of LinearRegression class present in the Scikit-learn library and is trained on the data (Treleaven et al., 2023). From the values of VWC of the media, the model is expected to predict the temperatures and these are written in a new column ‘Predicted_Temperature’. As any regression model, the presented model is evaluated based on root mean squared error (RMSE) and the coefficient of determination (R²), which represent its accuracy and the extent of explained variation, respectively. These metrics are then printed for reference and they capture.

Figure 28 : Actual vs Predicted Temperature

4.9 Summary of Findings

4.9.1 Key Insights from Data Visualization

The performance measurement of the data visualization identified various patterns of CO2 levels and their correlation with other factors. For instance, although the line plot depicted the trends of CO2 concentrations in the atmosphere, it especially pointed out areas of high concentrations. The histogram and the KDE plot helped to gain an understanding of how the distribution of CO2 is, whether it is high or low. Analysis of the CO2 levels using box plots exposed differences between particular devices, as well as problems with sensors’ calibration. The correlation heatmap of the current analysis also depicted the associations between CO2 and another variable while a scatter plot showed the association between height and CO2 level.

4.9.2 Key Insights from Time Series Analysis

In the case of CO2 level analysis, the elements of Time Series analysis gave fine-grained insight into temporal changes in the levels. The former was used to establish the trend which arose from the examination of the long-term tendencies and the latter was used to show the cyclic tendencies (Rathika and Pushparaj, 2023). It gave forecasts and information on trend and seasonality variables by the application of the ARIMA and SARIMA models. Moving averages and exponential smoothing organised trends and made it easier to do short-term forecasting. VAR and Prophet models looked into the interdependencies of the variables and delivered forecasts accompanied by uncertain ranges. CFT results in dynamic cross-correlations indicating that CO2 concentration interacts with other measured parameters.

4.10 Conclusion

4.10.1 Summary of Analysis

All the aspects related to the variables of the environment, including the levels of CO2, were highlighted in Chapter 3, with the use of data preprocessing, visualization and time series analysis. This faculty of analysis and the preprocessing steps facilitated ensured that this data was clean and consistent hence no erroneous results were likely to be produced. Graphical and graphical-based methods were used to display trends, distributions, and relations between variables and times series analysis gave information about patterns and forecasts through time.

4.10.2 Implications for Further Research

Based on the analysis it is possible to outline several directions for further research. For instance, other variables inside or outside the model could be tested, such as additional factors that might affect the concentration of CO2 in the environment to get an even better assessment of the environment. Higher-level models and duration of data could help to increase the accuracy of forecasts and unveil other patterns.

4.10.3 Practical Applications

The study is relevant for the practical assessment and regulation of environmental conditions defined and discussed in the work. Information about CO2 and fluctuations in its figures can effectively be used for developing measures aimed at the reduction of emissions and the improvement of air quality (Kyrpota, 2023). The insights developing from the analysis can also be used to improve the design of the structure of environmental monitoring systems and policies.

CHAPTER 5

CONCLUSION

5.1 Overview of the Study

The purpose of this work was to investigate the patterns of fluctuation of CO2 levels and their relations to other environmental factors and to use data visualization and time series analysis as methods of investigation. In particular, the investigation was directed at such factors as tendencies and cyclicity of CO2 content, as well as at the problem of CO2 prediction and its dependence on other parameters, for instance, temperature, and VWC. In this research work, which included data analysis on CO2 data, the following techniques were utilised: ARIMA, SARIMA, exponential smoothing, and cross-correlation analysis to reveal patterns.

5.2 Summary of Key Findings

5.2.1 Data Visualization Insights

Several significant findings relating to the characteristics of CO2 levels were made known at the data visualization stage of the study. Continuous line plots pointed to key temporal patterns, so fluctuations in high levels of CO2 were presented. Histogram and the Kernel Density Estimate plots helped in understanding the nature of the spread of the data with the densities of the CO2 levels shown to have a bell shape indicating occurrence in higher frequencies in certain scale ranges (Salam and Salam, 2020). Variations detected through the boxplots were related to concerns with the calibration and coherence of the employed measurement devices. Heat maps of correlations and scatter plots provided a better understanding of interactions between CO2 levels and the other drivers like temperature and VWC implying that there are more intricate correlations.

5.2.2 Time Series Analysis Insights

Analyzing the data by time series allowed finding out more detailed information on the temporal distribution of CO2 levels. Relative changes in CO _{2} concentrations have been identified in the course of analysis by trend detection which shows that they are either gradually increasing or decreasing in the long run. Seasonal decomposition broke the time series into trend, seasonality and remainders, and helped to recognize periodic patterns and seasonality. Thus, the models of ARIMA and SARIMA with components of trend and seasonality made possible the forecasting of conditions with high accuracy, while simple and smoothed methods of Winters’ model, moving averages and accuracy of short-term fluctuations. VAR model: Data interdependencies between CO2 and other variables were achieved in the model whereas uncertainty intervals of forecasts, which catered to strong seasonal effects were provided by the Prophet model. Results from cross-correlation highlighted aspects such as the lag effect of CO2 as well as the influence of one variable on the other.

5.3 Implications of Findings

5.3.1 Environmental Monitoring and Management

As a result, exploratory research of this type has important implications for environmental supervision and control. It is only possible to develop a specific approach to reduce the concentrations of CO2 and other greenhouse gasses in the Earth’s atmosphere if there is a clear understanding of the factors influencing the CO2 concentrations. The information that flows from the analysis will be useful in making policy shifts that will seek to tackle the concentration of CO2 in the atmosphere and climate change (Soheli et al., 2022). The analysis of fluctuations and trends in the levels of CO2 emissions can help stakeholders focus more on the specific areas where emissions may be reduced, as well as introduce measures to improve the sustainability of the environment.

5.3.2 Forecasting and Predictive Analytics

The used forecasting models such as ARIMA, the SARIMA and the Prophet models presented useful instruments for predicting future levels of CO2. Projections allow for avoiding situations where concentrations of CO2 rise higher and carrying out timely action. Fundamentally, the management of environmental interactions requires a forecast of seasonal fluctuations and long-term patterns so that fair solutions to unstable conditions and alteration can be established.

5.3.3 Data-Driven Decision Making

By reviewing data on the carbon dioxide concentration and other characteristics it is clear that the quantitative approach is crucial. When decision-makers apply analytical tools and presentation in the best way; it becomes easier to understand the current state of the environment and thus facilitate good decision-making. This kind of strategy also relates to responsiveness to new trends and therefore improves the adaptability of the organizational environment.

5.4 Recommendations for Future Research

5.4.1 Expanding the Dataset

There are several directions for further research that can be derived from the present study’s limitations: First, one can consider using a larger set of variables and a longer time series. The use of other pollutant data, meteorological conditions, and changes in land use might help in giving a more comprehensive situation of the factors that affect CO2 concentrations. If more extended data were available, it would be possible to consider more extended trends in the model and make forecasting more accurate and less sensitive to variations in values.

5.4.2 Advanced Modeling Techniques

Exploring other Multivariate Statistical Analyses like, Machine learning algorithms and Deep learning techniques can improve the forecast accuracy and subsequently uncover other intricate relationships in CO2 data (Bersani et al., 2022). Methods such as neural networks and ensemble methods might have enhanced predictive performance and detail about non-linear relationships between the variables.

5.4.3 Regional and Temporal Variability

Analyzing the variations in the geographical and temporal CO2 profiles could also give more information about the conditions concerning the region, and the factors that affect the levels of CO2. Comparing the data from the different geographical areas and periods would then assist in establishing specific area trends and this would assist in proper environmental management strategies.

5.4.4 Integration with Policy and Regulation

Future research should, therefore, seek to coordinate its results with the policies and regulations in place. In this way, by connecting the collected data with the current environmental policies and legislations, the researchers can help to enhance the effectiveness of the stated strategies regarding CO2 emissions and climate change.

5.5 Practical Applications

5.5.1 Policy Development

The findings of this study are useful for policy-making. Studying the tendencies and fluctuations in CO2 emission allows to creation and initiate specific policies and action plans aimed at reducing emissions and improving the air quality. The information obtained from the forecasting models may be used to craft sensitive policies for seasonal shifts and establish trends.

5.5.2 Environmental Monitoring Systems

The study shows that communities ought to enhance strong environmental check mechanisms. The methods of monitoring should involve the use of enhanced analytical methods and should produce timely and relevant information for decision-making. This study’s findings suggest ways in which monitoring processes of the environment can be made more efficient to monitor and analyse information concerning CO2 emissions.

5.5.3 Public Awareness and Engagement

To increase people’s consciousness about the relevance of CO2 emissions and the effects they have on the surroundings, it is vital to ensure that the general population is informed. It is suggested that the results of this study should be utilised in raising public awareness about the need to measure the concentration of CO2 and link personal contributions to the problem and its solution (Kashyap and Kumar, 2021). Interacting with the public using the data most appreciate and understand statistics and figures can go a long way in creating awareness and support for environmental causes.

5.6 Final Thoughts

In conclusion, this study gives an ample description of the variations in CO2 levels and their dependencies on environmental factors. It is an important discussion of mathematical thinking, with the use of data visualisation and time series analysis providing a richer understanding of CO2 trends and, in particular, seasonality and forecasting. The research underlines the necessity of the usage of numeric results to make conclusions and shows more practical aspects of environmental control and the creation of related policies. The literature should seek to draw larger samples, employ more elaborate statistical methods, and harmonize the results with policies for the effective management of environment-related issues. Due to ongoing study and use of analytical methods, one can improve his/her knowledge of environmental changes and thus work toward creating a sustainable environment.

Top Assignment Samples

The Role of Digital Technologies for the Growth of Entrepreneurial Firms	Market Analysis Of Hemp Heros For Expanding Hemp Products In Pet Sector
Comparison of Autocratic And Democratic Leadership Models	PLACEMENT REPORT ON M.V.T LIMITED T/A VENISONS
Impact of Work-Life Balance on Employee Retention in the Events Industry	Environmental Management And Waste Frame Directive Legislation in UK
AI and Big Data Impact on Tesco’s Supply Chain Management In UK Food Retail Sector	Role of MNCs In Global leadership and Corporate Citizenship
Beegrip International Consultancy Project Report	Impact Of Cyber Physical Frameworks or CPS On International Businesses
Global Strategy and Sustainability For ITunes	Type 2 Diabetes Mellitus (T2DM) Condition Analysis
Business Ethics Overview And Recommendations for Boohoo Company	International Business Trends Emerging Out of Covid-19
Wellness at Work: A Critical Examination of Employee Mental Health and Well-Being at Amazon	New Creative Mini Campaign for Innocent Drinks Targeting GenZ Audience

Reference

Agustina, A., Cahyania, M.O. and Syaoqibihillah, M., 2024. Data Privacy and the Law: Balancing Security and Individual Rights. Law Studies and Justice Journal (LAJU), 1(1), pp.15-24.

Bersani, C., Ruggiero, C., Sacile, R., Soussi, A. and Zero, E., 2022. Internet of Things approaches for monitoring and control of smart greenhouses in Industry 4.0. Energies, 15(10), p.3834.

Bilotta, G., Genovese, E., Citroni, R., Cotroneo, F., Meduri, G.M. and Barrile, V., 2023. Integration of an innovative atmospheric forecasting simulator and remote sensing data into a geographical information system in the frame of agriculture 4.0 concept. AgriEngineering, 5(3), pp.1280-1301.

Bouali, E.T., Abid, M.R., Boufounas, E.M., Hamed, T.A. and Benhaddou, D., 2021. Renewable energy integration into cloud & IoT-based smart agriculture. IEEE Access, 10, pp.1175-1191.

Chen, T., Ma, J., Liu, Y., Chen, Z., Xiao, N., Lu, Y., Fu, Y., Yang, C., Li, M., Wu, S. and Wang, X., 2022. iProX in 2021: connecting proteomics data sharing with big data. Nucleic acids research, 50(D1), pp.D1522-D1527.

Doshi, J., Patel, T. and kumar Bharti, S., 2019. Smart Farming using IoT, a solution for optimally monitoring farming conditions. Procedia Computer Science, 160, pp.746-751.

Fenoglio, E., Pithadia, H. and Treleaven, P., 2023. Federated Computing. Available at SSRN.

Hassan, M., Hussein, A., Nassr, A.A., Karoumi, R., Sayed, U.M. and AbdelRaheem, M., 2024. Optimizing Structural Health Monitoring Systems Through Integrated Fog and Cloud Computing Within IoT Framework. IEEE Access.

Henderson, A., Yakopcic, C., Colter, J., Harbour, S. and Taha, T., 2023, October. Blockchain-Enabled Federated Learning with Neuro

Hung, C.H., Fanjiang, Y.Y., Wang, D.J. and Chen, K.C., 2023. Contactless Water Level Detection System for Pickling Barrels Using Image Recognition Technology. IEEE Access.

Irawan, E. and Muzakir, A., 2022. Sistem Pengendali Keamanan Sepeda Motor Berbasis IoT (Internet of Things) Menggunakan Smartphone Android. Journal of Information Technology Ampera, 3(2), pp.148-158.

Kalinaki, K., Malik, O.A., Yahya, U. and Lai, D.T.C., 2024. Federated learning challenges and risks in modern digital healthcare systems. In Federated Learning for Digital Healthcare Systems (pp. 283-300). Academic Press.

Kashyap, B. and Kumar, R., 2021. Sensing methodologies in agriculture for soil moisture and nutrient monitoring. IEEE Access, 9, pp.14095-14121.

Keswani, B., Mohapatra, A.G., Mohanty, A., Khanna, A., Rodrigues, J.J., Gupta, D. and De Albuquerque, V.H.C., 2019. Adapting weather conditions based IoT enabled smart irrigation technique in precision agriculture mechanisms. Neural Computing and Applications, 31, pp.277-292.

Kyrpota, F., 2023. Development of Automated Environmental Control System for Portable Greenway Section.

Madushanki, A.R., Halgamuge, M.N., Wirasagoda, W.S. and Syed, A., 2019. Adoption of the Internet of Things (IoT) in agriculture and smart farming towards urban greening: A review. International Journal of Advanced Computer Science and Applications, 10(4), pp.11-28.

Madushanki, A.R., Halgamuge, M.N., Wirasagoda, W.S. and Syed, A., 2019. Adoption of the Internet of Things (IoT) in agriculture and smart farming towards urban greening: A review. International Journal of Advanced Computer Science and Applications, 10(4), pp.11-28.

Maraveas, C., Piromalis, D., Arvanitis, K.G., Bartzanas, T. and Loukatos, D., 2022. Applications of IoT for optimized greenhouse environment and resources management. Computers and Electronics in Agriculture, 198, p.106993.

Marcu, I., Suciu, G., Bălăceanu, C., Drăgulinescu, A.M. and Dobrea, M.A., 2019, October. IoT Solution for Plant Monitoring in Smart Agriculture. In 2019 IEEE 25th International Symposium for Design and Technology in Electronic Packaging (SIITME) (pp. 194-197). IEEE.

Marques, P., Manfroi, D., Deitos, E., Cegoni, J., Castilhos, R., Rochol, J., Pignaton, E. and Kunst, R., 2019. An IoT-based smart cities infrastructure architecture applied to a waste management scenario. Ad Hoc Networks, 87, pp.200-208.

Ming, F.X., Habeeb, R.A.A., Md Nasaruddin, F.H.B. and Gani, A.B., 2019, February. Real-time carbon dioxide monitoring based on iot & cloud technologies. In Proceedings of the 2019 8th International Conference on Software and Computer Applications (pp. 517-521).

Mohammed, R.S., Malhotra, R., Shamout, M.D., Pithadiya, B., Patil, S. and Sangve, S.M., 2023, December. High-Accuracy Crop Yield Estimation Through IoT and Remote Sensing. In 2023 IEEE International Conference on Paradigm Shift in Information Technologies with Innovative Applications in Global Scenario (ICPSITIAGS) (pp. 225-230). IEEE.

Naresh, V.S., Pericherla, S.S., Murty, P.S.R. and Reddi, S., 2020. Internet of Things in Healthcare: Architecture, Applications, Challenges, and Solutions. Computer Systems Science & Engineering, 35(6).

Pais, V., Rao, S., Muniyal, B. and Yun, S., 2024. FedICU: a federated learning model for reducing the medication prescription errors in intensive care units. Cogent Engineering, 11(1), p.2301150.’

Patrizi, G., Bartolini, A., Ciani, L. and Catelani, M., 2022. Failure analysis of a smart sensor node for precision agriculture. management, 1, p.3.

Patrizi, G., Bartolini, A., Ciani, L., Gallo, V., Sommella, P. and Carratù, M., 2022. A virtual soil moisture sensor for smart farming using deep learning. IEEE Transactions on Instrumentation and Measurement, 71, pp.1-11.

Pithadiya, B.H., Parikh, H.N., Pandya, H.N. and Vyas, D., 2021, November. IoT based Automation of Public Garden and Botanical Garden. In Journal of Physics: Conference Series (Vol. 2089, No. 1, p. 012062). IOP Publishing.

Ramson, S.J., León-Salas, W.D., Brecheisen, Z., Foster, E.J., Johnston, C.T., Schulze, D.G., Filley, T., Rahimi, R., Soto, M.J.C.V., Bolivar, J.A.L. and Malaga, M.P., 2021. A self-powered, real-time, LoRaWAN IoT-based soil health monitoring system. IEEE Internet of Things Journal, 8(11), pp.9278-9293.

Ratanpara, J. and Pithadia, N., Efficient Network Management through Python Scripting.

Rathika, P.D. and Pushparaj, A., 2023, February. Privacy Preservation Using Federated Learning for Credit Card Transactions. In 2023 International Conference on Intelligent Systems for Communication, IoT and Security (ICISCoIS) (pp. 398-403). IEEE.

Salam, A. and Salam, A., 2020. Internet of things for environmental sustainability and climate change. Internet of Things for sustainable community development: Wireless communications, sensing, and systems, pp.33-69.

Soheli, S.J., Jahan, N., Hossain, M.B., Adhikary, A., Khan, A.R. and Wahiduzzaman, M., 2022. Smart greenhouse monitoring system using internet of things and artificial intelligence. Wireless Personal Communications, 124(4), pp.3603-3634.

Spoorthi, M. and Gururaj, H.L., 2024. on Federated Applications Recent. Federated Learning Techniques And Its Application In The Healthcare Industry, p.69.

Treleaven, P., Barnett, J., Brown, D., Bud, A., Fenoglio, E., Kerrigan, C., Koshiyama, A., Sfeir-Tait, S. and Schoernig, M., 2023. The future of cybercrime: AI and emerging technologies are creating a cybercrime tsunami. SSRN.

Vanus, J., M. Gorjani, O. and Bilik, P., 2019. Novel proposal for prediction of CO2 course and occupancy recognition in intelligent buildings within IoT. Energies, 12(23), p.4541.