Greenhouse gases constitute the primary source of temperature elevation, with methane playing a significant role, mainly produced in Syria through the decomposition of organic matter in oxygen-free environments. Approximately 40 % of methane is naturally produced, while the remaining 60 % is of anthropogenic origin, primarily associated with agricultural practices, particularly livestock farming, waste processing and the coal and oil industry.

In addition to these gases, nitrous oxide, often referred to as “laughing gas,” is primarily produced through intensive agricultural practices such as fertilization, the combustion of fossil fuels, and some industrial processes. Despite methane and nitrous oxide being present in the atmosphere at lower concentrations compared to carbon dioxide, their contribution to global warming is significantly greater. For example, the warming potential of nitrous oxide is 265 times higher than that of carbon dioxide.

​It is indicated that approximately 75% of losses are attributable to extreme weather events. Climate change poses a threat of pushing an additional 100 million people into poverty by 2030. Another report from the World Bank, also funded by this institution, highlights that natural disasters have substantial and long-term effects on poverty(World Development Report 2016: Digital Dividends, n.d.).

The 2015 Paris Agreement establishes a global framework for constraining global warming to below 2°C, with a preferred target of 1.5°C, relative to pre-industrial levels (Grillakis et al., 2022). To achieve this temperature objective, nations strive to curb the increase in greenhouse gas emissions as soon as possible and subsequently implement rapid reductions, based on the best available science and considerations of economic and social feasibility (Everest Base Camp Trek | Everest Base Camp Trek Cost 2024-2025, n.d.).

A comprehensive review of literature was conducted concerning the utilization of neural network models and fuzzy logic across various environmental domains, including water treatment and weather forecasting. The results of the literature analysis revealed that the majority of AI-based prediction models have been employed in addressing issues related to water/wastewater treatment (55.7%) and air pollution (30.8%), whereas studies focusing on solid waste management comprise a smaller proportion (13.5%). Specifically, in the realm of air pollution, predictive models have been applied to forecast levels of ozone (O₃), nitrogen dioxide (NO₂), daily and/or hourly emissions of particulate matter (PM_2.5 and PM₁₀), sulfur dioxide (SO₂) and carbon monoxide (CO). Conversely, in the context of solid waste management, researchers have conducted numerous investigations aimed at modeling the weight of generated waste, the composition of solid waste, and the overall rate of waste generation (Yetilmezsoy et al., 2011).

The 2012 Doha meeting marked a continuation of the Kyoto Protocol, a legally binding global agreement wherein participating nations committed to reducing their carbon emissions. Despite this supposed commitment, a notable increase in carbon emissions has been observed among all G20 countries, with the exception of France and the United Kingdom. China, in particular, has seen CO₂ emissions surge by over 300% in the past 25 years (Tudor, 2016). As per the Doha Amendment to the Kyoto Protocol, Bahrain’s baseline level corresponding to 1990 stands at approximately 25.54 metric tons per capita, necessitating a maximum level of 20.96 metric tons per capita for 2020 (reflecting an 18% decrease relative to the baseline level) for Bahrain compliance. However, findings suggest that Bahrain is unlikely to achieve its presumed target (Tudor, 2016).

The effects of climate change on the dust phenomenon in the Zanjan region, northwestern Iran, were investigated utilizing a neural network (ANN) model extending until the year 2050. Particle concentrations were analyzed based on daily meteorological data from 2007 to 2018 and 1988 to 2018, directly from the city of Zanjan. The Long Ashton Research Station Weather Generator (LARS-WG 6.0) software was employed to estimate statistical errors and generate high-resolution synthetic time series climate datasets. Climate variables such as maximum and minimum temperature and precipitation were identified as predictive variables in the artificial neural network (Asadollahfardi et al., 2015). The results showed that the highest increase in PM10 levels occurred in May and July, while the lowest increase was observed in December, with monthly average of 84.85 and 50.54 µg/m3, respectively. The peak PM₁₀ concentration was projected for the year 2043, reaching 74.26 µg/m3. Furthermore, the study highlighted significant relationships between PM₁₀ concentrations and variables including minimum and maximum temperatures, as well as wind speed. Additionally, it was noted that pollutant levels increased with the enhancement of each atmospheric variable (Moghanlo et al., 2021).

Forecasting CO₂ emissions is crucial for selecting optimal strategies to mitigate CO₂ emissions effectively. The machine learning (ML) method employed in the IPCC model has demonstrated remarkable accuracy in predicting historical emissions, current emissions impacted by COVID-19, and future emissions. These expected results serve as valuable inputs for decision support systems aimed at formulating appropriate policies to reduce global carbon dioxide emissions. Future research endeavors may consider incorporating additional exogenous variables that influence CO₂ emissions to enhance prediction accuracy further. Furthermore, predicting CO₂ emissions influenced by COVID-19 using AI-driven machine learning methodology is an accurate process (Meng & Noman, 2022).

In light of global warming and the energy crisis, the Chinese government has set goals to achieve carbon peaking and carbon neutrality. With the rapid advancement of machine learning, several advanced machine learning algorithms have been applied to manage and predict carbon emissions due to their high efficiency and accuracy. The current state of machine learning applied to carbon emission forecasting was thoroughly examined (Zhao et al., 2023). The findings show that machine learning has emerged as a significant subject in the field of carbon emission forecasting models. The primary models for carbon emission forecasting predominantly rely on backpropagation neural networks, support vector machines, long short-term memory neural networks, and stochastic neural networks (Zhao et al., 2023).

A study conducted in Saudi Arabia utilized three powerful and efficient intelligent modeling methods: feed-forward neural network (FFNN), fuzzy network-based inference system (ANFIS), and long short-term memory (LSTM). These methods were applied within a network of FFNN, ANFIS, and LSTM. The results showed that the average error values for each method were 19.78, 20.89505, and 15.42295, respectively, while the average correlation coefficients were 0.990985, 0.98875, and 0.9945, respectively (Nassef et al., 2023). Each model was separately applied to predict CO₂ emission values. In order to leverage the benefits of intelligent modeling tools, the final predicted value was determined as the average (combined) value of the outputs from the three model. To assess prediction accuracy, the group underwent validation with new measurements of the outputs from the three models. In 2021, the results revealed a calculated error rate of 6.8675% and an accuracy rate of 93.1325%, indicating a high level of accuracy for the model (Nassef et al., 2023).

Residual amounts of carbon dioxide, a phenomenon influenced by greenhouse gas emissions resulting from digital technology, were estimated in London. The results illustrate the significant accuracy of neural network estimation and emphasize the necessity for further evidence regarding the trade-off between greenhouse gas emissions resulting from technological advancements and the energy and resource efficiency gains achievable through AI. Effectively leveraging the opportunities presented by AI to address global climate change while mitigating associated risks necessitates swift, evidence-based, and efficient action, ultimately leading to a successful strategy (Cowls et al., 2023).