Enhancing climate forecasting with AI: Current state and future prospect

Introduction

The atmosphere around us is reflected in the weather. One aspect of a natural occurrence that keeps the atmosphere in balance is the weather (Subhajini, 2018). The elements of the weather include temperature, humidity, wind speed, precipitation, evaporation, air pressure, vapour pressure, length of sunshine, sea level, visibility, etc. In addition, weather forecasting affects our daily lives highlighting it as a significant scientific problem to be explored. It is one of the crucial strategy to prevent adverse weather conditions. Forecasting the weather is important for making decisions in a variety of areas, including business, energy management, tourism, agriculture, and the health of people and animals (Gautam & Bedi, 2015). Weather forecasting has been one of the world’s most difficult challenges in experimentation and technology throughout the past century (Salman, Kanigoro, & Heryadi, 2015). For a long time, environmental change has drawn a lot of interest because of the sudden shifts that occur. This study utilized a systematic bibliometric methodology to analyze peer-reviewed literature from the past two decades on the application of Artificial Intelligence (AI) in climate forecasting. The literature search was carried out on December 22, 2023, using the Scopus and Web of Science databases. The search strategy included the keywords (“Artificial Intelligence” OR “AI”) AND (“Weather forecasting” OR “Climate forecasting”), restricted to articles published in English between 1999 and 2023.

Inclusion criteria involved peer-reviewed articles focusing on AI applications in climate or weather forecasting. Exclusion criteria included duplicate articles, non-English publications, and non-peer-reviewed articles.

However, due to certain limitations in improving weather forecasting implementation, it is challenging to predict the weather here and now with precision (Hussain et al., 2018). A major part of meteorology is weather forecasting (Fradkov, 2020). Accurate prediction-making is one of the main problems that meteorologists face globally. Weather alerts are crucial because they are used to safeguard lives and property. Forecasts based on temperature, wind, humidity, and outlook are crucial for farmers and, consequently, for traders in product markets. Temperature forecasts are used by utility providers to assess demand for the upcoming few days (Singhroul & Agrawal, 2021). Estimates can be used to schedule activities around periods of wind chill, heavy rain, or snowfall, as these weather conditions severely limit outside activity. They may also be utilized to weather and be ready for these circumstances. Without reliable weather forecasts, individuals may find themselves unprepared for hazardous conditions and endanger themselves or worse (Abuella & Chowdhury, 2018).

The process of creating weather forecasts involves obtaining information about the current state of the atmosphere in a certain area and using the weather to anticipate future changes to the atmosphere (Awasthi et al., 2023). Individual input is still required to select the best prediction model and create the forecast (Adam, 1991). Meteorologists predict the weather using a variety of techniques. In the early 20th century, Lewis Fry Richardson suggested numerical weather prediction. The first practical application of numerical weather prediction came in 1955 with the introduction of programmable electronic computers. Near the Earth’s surface, air pressure, temperatures, wind speeds, wind directions, humidity, and rainfall are all monitored by trained meteorologists and automated weather stations and then used in different climate models (Khan et al., 2022). Determining how to represent the weather by utilising a vast quantity of meteorological data is one of the issues in weather forecasting. For this purpose, an analysis of different data mining approaches is conducted. Users may classify, evaluate, and combine the relationships revealed in data from several dimensions or points of view with the use of data mining tools (Akhter et al., 2019). In data mining, terms like classification, learning, and prediction are frequently employed. Classification is one method used in data mining (machine learning) to predict aggregate participation for information circumstances. To predict whether the weather will be “sunny,” “rainy,” or “cloudy” on a certain day, for instance, classification (Awasthi et al., 2023). Analytical approaches and artificial intelligence methods are the two basic categories in which the methods created for load forecasting are analyzed. Time series analysis, regression techniques, the similar day method, least square estimation (LES), and wavelet transform (WT) are often employed as analytical techniques in the literature. These AI techniques include ant colony optimization (ACO), fuzzy logic (FL), support vector machines (SVM) (Aylak, 2021), genetic algorithms (GA), particle swarm optimization (PSO), and artificial neural networks (ANN) (Bochenek & Ustrnul, 2022). To achieve good forecasting, load demand-affecting factors need to be carefully identified and taken into account in forecasting studies. Consumption region and independent variables such as day of the week, and sociological, climatic, demographic, and economic conditions can be used to characterize these parameters. Statistical and economic circumstances are typically taken into account in long-term load forecasting studies (Bauer, Thorpe, & Brunet, 2015). Studies also make use of climatic variables including wind, precipitation, and humidity in addition to temperature data. Studies that exclusively use historical load data also exist (Bonavita & Laloyaux, 2020).

Implications for future research and practice

The findings of this study have significant implications for future research and practice. By identifying the most influential articles, authors, and research themes, the study provides a roadmap for future investigations. It highlights the need for developing universal AI models, enhancing model accuracy with explainable AI techniques, and integrating region-specific forecasts. This will aid in better planning and decision-making across various sectors, contributing to the broader goal of climate resilience and sustainability.

This research is justified by its potential to advance the field of climate forecasting through the application of AI, addressing critical gaps in the literature, and providing actionable insights for future research and practical applications. The enhanced accuracy and early warning capabilities offered by AI can significantly contribute to mitigating the impacts of climate change, particularly in vulnerable regions. Below Figure 2 shows the structure of study.

Figure 2. Structure of study.

Source: Prepared by the researcher(s).

A systematic bibliometric methodology was employed, analyzing peer-reviewed literature from the past two decades ( Figure 2).

Methods

The aim of this study is to understand the scope of conducting research in the field of artificial intelligence and weather forecasting. The science mapping analysis and performance analysis were carried out using Bibliometrix R. Artificial intelligence predicts climate condition (Chen et al., 2023a,b,c). A variety of indices were used for the investigation via Performance Analysis, including citations, annual scientific productivity, highly productive nations, and highly cited publications and authors. Secondly, scientific mapping is used where different clusters that were thematically related were examined using the co-word co-occurrence analysis. A cluster is characterized as a collection of nodes or keywords that share many connections and are more closely grouped together than the other groups (Chung, Gray, & Mannino, 1998). Whereas the betweenness of centrality gauges a node’s significance by counting the shortest pathways through it and indicating its influence over the network (Cifuentes et al., 2020). Strongly correlated keywords were identified by the clustering network (Cobo et al., 2011). Figure 3 indicates PRISMA study for review of research. Lastly, to discover the research themes the overall data has been plotted on a strategic diagram. where a set of contemporary trends and related themes were extracted through analysis of the strategic diagrams. The most reliable scientific data regarding the impact of health interventions are obtained through systematic reviews that integrate meta-analyses of randomized clinical trials. Nevertheless, certain meta-analyses and systematic reviews are unclear and omit crucial details, especially when it comes to the methodology and findings (Cowls et al., 2021). The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) declaration aims to increase transparency. Authors and researchers across the world have been using the PRISMA statement to organize, draft, and publish systematic reviews and meta-analyses. To assess the level of development and relevance of the terms, respectively, the density and centrality parameters were examined with the help of artificial intelligence (Fente & Singh, 2018).

Figure 3. PRISMA selection criteria.

The study screened 455 articles from Scopus and Web of Science databases, following the PRISMA flow diagram ( Figure 3).

Existing relation between artificial intelligence and climate forecasting

Artificial intelligence (AI) offers the potential to be significant in tackling difficulties associated with weather change. The application of AI to lower Greenhouse Gas (GHG) emissions, especially in the electric power industry, is the focus. By addressing low-hanging fruit in industries like electricity, transportation, agriculture, and buildings, AI can help tackle weather change (Davenport & Diffenbaugh, 2021). Weather change addresses data and technical difficulties and highlights the potential uses of AI in lowering emissions GHG (Devaraj et al., 2021). Meteorology relies heavily on weather predictions. Making a precise forecast is one of the most difficult things to stand up to meteorologists in any part of the globe. Given that weather warnings are used to protect people and property, they are essential. To increase the models’ accuracy, researchers have put out a variety of weather forecasting models. Precise prediction may not maybe because of noise and inaccurate or missing data. Real-world data tends to be noisy and unpredictable (Dong et al., 2023). Weather prediction systems have used a variety of techniques, such as neural network-based algorithms that use backpropagation. A neural network technique used in machine learning for feature extraction is called a stacked auto-encoder (Dupuy et al., 2021). To accomplish dynamic data interaction, or the link between expected outcomes and actual data in a dynamic context, a stacked auto-encoder neural network might be useful. There have been several noteworthy attempts in the last ten years to use statistical modeling, including machine learning systems, to address the problem of weather forecasting. favorable outcomes (Elwell & Polikar, 2011). In the last ten years, artificial intelligence has emerged, and to research coastal dynamics, machine learning models have been used. Traditional methods of predicting the weather include constraints, but a novel AI-driven strategy utilising deep learning and machine learning techniques exhibits the potential to deliver more fast and more accurate weather predictions (Era, Rahman, & Alvi, 2023). For weather forecasting, several artificial intelligence algorithms are frequently utilised. Examining historical meteorological data, predicting future weather conditions using trends and supervised machine learning methods like Random Forest and Support Vector Machines may be used to identify trends as well as neural systems. Weather forecasting has made use of Random Forests as it can handle high-dimensional data and non-linear correlations (Ling et al., 2022). Large data sets and intricate patterns may be analysed by AI systems to generate extremely precise weather forecasts. This precision is essential for proactive network management and planning (Leal Filho et al., 2022). Real-time weather updates may be obtained via AI-enhanced systems, enabling network managers to take prompt action to reduce disturbances in the network (Kumar & Palanisamy, 2020).

Artificial intelligence-driven weather forecasting can help communication networks become more equipped to handle extreme weather occurrences making sure they continue to function in urgent circumstances (Gowthamy et al., 2019). By forecasting localised meteorological conditions, AI can help with resource allocation, ensuring that resources are allocated to locations where they are most required (Abdel-Kader and Salam, 2021). Artificial Intelligence can analyse unstructured data sources, like social media, and offer a more detailed picture of the present weather (Hedar et al., 2021).

Medium-range weather forecasting is another important factor that is influenced and predicted using AI (Hensengerth, 2024). The AI-based weather forecasting models nevertheless rely on analytical products produced by the classic NWP system’s data assimilation process to make forecasts (Huntingford et al., 2019). Conventional weather forecasting uses millions of equations to mimic detailed atmospheric processes through complex physical models. Accurate representation of atmospheric phenomena is the goal of these models. These models consider information from weather stations, radar, satellites, and other sources, along with a variety of parameters like temperature, humidity, wind speed, and cloud cover (Pathak et al., 2022). Artificial Intelligence (AI) has multiple applications in predicting, including natural language processing, deep learning, and machine learning. AI is superior to human forecasters in several ways. It can handle big and varied datasets, spot patterns and trends, automate procedures, and adjust to shifting circumstances and uncertainties to improve forecasting models’ accuracy, dependability, simplicity, and flexibility Neema and Ohgai (2010). These forecasts use computational methods and scientific information to help us anticipate future weather phenomena, such as rain, snow, or sunshine (Grenier et al., 2023). With the advent of more affordable sensing units and improved connectivity, the Internet of Things (IoT) is becoming more accessible. As a result, the variety of gadgets and devices that can provide valuable real-time weather data is expected to expand significantly (Bi et al., 2023). Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) were used in 2020 to estimate the daily temperatures for the next three days, producing a model. The five temperature categories used were “Cold,” “Cool,” “Normal,” “Warm,” and “Hot.” Data The accuracy of the results was influenced by the size and distribution of the dataset used. (Chen et al., 2023a,b,c) Harnessing AI’s potential in weather forecasting is exemplified by IBM’s acquisition of The Weather Company. This acquisition empowered IBM to generate highly accurate weather forecasts using the power of AI. Such forecasts hold significant value for businesses in the communication sector as they aid in informed decision-making (Kaur et al., 2018). However, despite the groundbreaking nature of this approach, concerns arise regarding the scalability and adaptability of AI across diverse meteorological scenarios (Khotanzad et al., 1997). Weather predictions become less precise as the prediction range increases and the lack of knowledge about atmospheric dynamics. It’s easy to understand why perfect weather forecasting matters when you consider the benefits (Kim et al., 2022). These forecasts won’t stop a tornado, storm or flood from happening, but they can help us plan for them. We’ve come a long way (Laloyaux et al., 2022).

Artificial intelligence (AI) is emerging as a potent tool in weather forecasting that could transform our knowledge and ability to predict weather patterns. AI has the potential to greatly increase our understanding of weather change and provide new insights into intricate weather systems. Additionally, AI is playing a crucial role in all-encompassing answers to weather catastrophes by offering more efficient, sustainable, and environmentally friendly weather forecasting technologies (Chen et al., 2023a,b,c). However, the use of AI in weather-related fields raises ethical issues like discrimination, unfair bias, and opaque decision-making, which are consistent with larger worries about AI applications. AI additionally highlights the significance of taking computation-intensive carbon footprint into account and the necessity of scientific proof to balance greenhouse gas emissions against increases in energy and resource efficiency (Kıran et al., 2012). There is an increased amount of research in the field of artificial intelligence and weather forecasting. However, there is a scarcity of bibliometric research in this area exploring the scope of conducting research in the field of artificial intelligence and weather forecasting (Singh et al., 2022). Bibliometric analysis will serve as a knowledge base for further researchers by providing all the necessary information associated with this area and what are the various domains associated with it. As above following work has already been done. Table 1 showing current state of work and future scope. Study include in Table 1 are taken from period 1999 to 2023. Reason behind to select the study to find out development of AI in climate forecasting and it future prospectus. Many researchers have developed a keen interest in the field of Artificial intelligence and weather forecasting (Martínez et al., 2022). Consequently, a bibliometric study was carried out to comprehend the present research trends in the field of Artificial intelligence and weather forecasting as well as to investigate the different other research domains connected to this field (Martín-Vázquez, Aler, & Galván, 2017).

Analysis publication of AI and climate forecasting

Publication trend

Figure 4, represents the publication trends in artificial intelligence and weather forecasting. The scientific output is derived from the past four decades i.e.,1984 to 2023. The publication analysis is divided into four distinct periods, characterized by its production and significance of the articles. The study select on the based of artificial intelligence development in climate forecasting from initial period when role of AI included in climate forecasting.

Figure 4. Annual Scientific Production of Artificial Intelligence and Weather Forecasting.

Source: Prepared by Author from R studio.

The first stage, the initial development stage (1984-1993), experienced a very slow and negligible growth. The total number of publications in this decade were only two. During this period researchers were trying to establish the relationship between computers and artificial intelligence. Zadeh (1984) has explained the use of artificial intelligence in computers with the example of fuzzy application. The fuzzy is nothing but an example showing how computers can think like a human along with its demerits. The second article in this also shows the usefulness of artificial intelligence in air traffic control using satellites. Artificial intelligence can be used in different aspects of day-to-day use like weather predication and information sharing (McGovern et al., 2019).

The second stage (1994-2003) has also witnessed a very a smaller number of publications where articles were published in different fields including weather forecasting, cloud classification, environmental science and climatic variation. In this decade various automated working computing programs were introduced by various authors. Authors further acknowledge the requirement of artificial intelligence in weather prediction (Medhekar, Bote, & Deshmukh, 2013).

The third phase of this topic lies between 2004 to 2013 and was characterized by the global adoption of artificial intelligence and its implications. According to scientific outcomes, a total no of 18 articles were published in these ten years. This phase can be considered as the introductory phase of the topic/relationship between artificial intelligence and weather predictions. In this phase, the relationship between artificial intelligence has been established in various aspects of research (Mehmood et al., 2021)

The last and the fourth stage (2014-2023) signifies the promotion and innovation, where the highest publication took place compared to last three decades. This stage contains 88 percentage of the articles used for the study which is 193. This stage consists of artificial intelligence and its various used in weather prediction.

The findings indicate significant growth in AI applications for climate forecasting, particularly from 2014 to 2023. The annual scientific production of AI and weather forecasting is depicted in Figure 4.

Most influential articles on artificial intelligence and weather forecasting

The value of articles can be seen in their citation scores. So, the analysis of the most influential article is also evaluated on the same phenomena, although the latest articles can face a disadvantage in this. Figure 5 shows the most cited and impactful articles in the area of artificial intelligence and weather prediction. With 661 citations, “Incremental Learning of Concept Drift in Nonstationary Environments” by (Muñoz-Leiva et al., 2012) is the most influential. In the article the author introduced an ensemble of classifier-based approach which can be utilized for incremental learning of concept drift, characterized by nonstationary environments. The article has also scored 42 total citations per year. With 436 citations, (Monteiro et al., 2022) conducted the second influential article “Solar photovoltaic generation forecasting methods: A review” on this theme. The study focuses on the concept of solar photovoltaic generation and its implications in power generation. The author reviewed various forecasting systems with significant information to design an optimal solar photovoltaic system (Nascimento et al., 2022).

Figure 5. Most prolific authors of artificial intelligence and weather forecasting.

Source: Prepared by Author from R studio.

The most influential articles on this topic, based on citation scores, are illustrated in Figure 6.

Figure 6. Most influential articles on artificial intelligence and weather forecasting.

Source: Prepared by Author from R studio.

Co-word co-occurrence analysis

Co-word analysis helps in identifying various factors and the global dynamics of the field under the study. Further, it uses a large knowledge base to extract various patterns of research trends. The cluster wise analysis provides the link between the topics within the cluster based on the betweenness and closeness of centrality where it will reflect the various trending topics in that field of study (Navadia et al., 2017). The Co-word co-occurrence analysis helps in exploring the various active themes of research as well as sub fields for the area under study (Nusrat & Jang, 2018). The author’s keyword in the study have been classified into four cluster based on the co-occurrence network. Figure 6 represents the interrelationship among the authors keywords and their significance as direct themes. The clusters selected on based of artificial intelligence role in climate forecasting and how many research paper published on AI on climate forecasting (Pattnayak et al., 2023).

Cluster 1: Artificial Intelligence and numerical weather prediction

The Figure 6 showing AI used in weather prediction. This cluster is the largest cluster related to artificial intelligence and weather forecasting where it consists of 52.9% of the key terms. The artificial intelligence algorithms are being one of the explicit approaches being utilised by researchers for weather forecasting basis where they provide best possible probabilistic forecast (Harder et al., 2022). Where the deep learning error correction methods effectively improve the prediction results of numerical weather prediction (Penny et al., 2022). The parameter of betweenness of centrality highlights that machine learning, artificial intelligence, numerical weather prediction, deep learning and forecasting are few of the central nodes among all the others keywords pertaining to this cluster. In this cluster, most closely related literatures are (Rasel, Sultana, & Meesad, 2018), (Rahayu et al., 2020), (Lam et al., 2022), (Riordan & Hansen, 2002), (Sallehuddin et al., 2007), (Sawale & Gupta, 2013), (Sedighi, 2016), (Sengoz et al., 2023), (Agrawal et al., 2023), (Sobri, Koohi-Kamali, & Abd Rahim, 2018), (Jiang et al., 2022),( Daniele & Sciacca, 2021), (Torres et al., 2021), (Yigitcanlar et al., 2021), (Trájer et al., 2022), (Yang & Wang, 2017), (Zaman et al., 2017) and (Ahmad & Kumar, 2024) where they represent the fundamental research in the field.

A co-word co-occurrence analysis highlights key themes and their interrelationships ( Figure 7).

Figure 7. Co-word mapping.

Source: Prepared by Author from R studio.

Cluster 2: Neural Networks and Numerical weather prediction

This cluster consist of 35.29% of key terms, where it explores the fundamental literature related to neural network and numerical weather prediction. The climate forecasting plays a vital role in many areas of human life. Estimating data for long-term efficient and effective results is very important, where predicting climate parameters based on AI, neural network and deep networks can optimise the data analysis and forecasting while simultaneously reducing errors (Yang et al., 2023). The artificial intelligence methods produce better results for sub seasonal to seasonal prediction (Mouatadid et al., 2023). Wherein, the scholars are suggesting the use of AI, neural networks, and numerical weather prediction to be used over conventional numerical forecast models for repeated initialisation of short-term forecast (Srivastava, Rastogi, & Kaushik, 2021). The keywords neutral network, numerical weather with prediction and statistical forecasting are central nodes among all the keywords. In this cluster, most closely related literature are (Stein, 2020), (Stuart et al., 2022), (Ravuri et al., 2021), (Ahmad, Zhang, & Yan, 2020), (UNFCCC, 2016), (Vitart et al., 2022).

Cluster 3: Feature selection and feature extraction

This cluster consists of 5.88% of the key terms, where it mainly explores the literature related to feature selection and feature extraction. (Vuyyuru, Rao, & Murthy, 2021) demonstrate that deep feature extraction, as opposed to feature selection, can enhance forecasting accuracy. Similarly, it is purported that automatic feature extraction methods allow deep learning (DL) algorithms to capture the inherent non-linear features, which permits them to handle big data (Wang & Huang, 2023). Wherein, the computational methods that are now in use do not investigate which climatic parameter or parameters have the greatest influence on achieving the best predicting performance. However, researchers suggest that various feature selection techniques can be utilised to identify the ideal subset of features (Wang & Mazharul Mujib, 2017). The keywords feature selection and feature extraction are the only central nodes in this cluster, where the most closely associated literature are (Wilde, 1994), (Wu, Wu, & Zhu, 2019), (Zhang & Chen, 1983), (Zhao et al., 2022).

Cluster 4: Learning (artificial intelligence) and power engineering computing

The literature on artificial intelligence and power engineering computers is the primary focus of this cluster, which comprises 5.88 percent of the key phrases. Numerous scholars have examined the various important advances in artificial intelligence and machine learning techniques (Gao et al., 2022). Wherein, Yang (2017) observed that the majority of these studies related to artificial intelligence and machine learning focuses on crop yield and soil properties prediction. Additionally, it is implied that the use of unmanned aerial vehicles, artificial neural networks, support vector machines, and remote sensors are rather common in weather prediction particularly in the field of agriculture (Rahim et al., 2023). In a similar vein, (Arora et al., 2023) offer a methodical and analytical analysis of the forecasting techniques, concentrating mostly on the metaheuristic and machine learning approaches. In this cluster, most closely related literature are (Xiao, Bennett, & Armstrong, 2002), (Dziejarski, Krzyżyńska, & Andersson, 2023).

Thematic analysis

To assess the current state of the themes, aggregation of the data is synthesized and represented on a strategic diagram ( Figure 7). The most prominent themes for the contemporary time frame are visualized using this diagram. The strategic diagram positions these themes according to the dimensions of centrality and density. Centrality gauges the degree of interaction within the networks; while, density measures the internal strength among all the keywords in the network (Hao, Singh, & Xia, 2018).

The thematic analysis revealed several prominent themes in contemporary research, visualized in a strategic diagram ( Figure 8).

Figure 8. Thematic analysis.

Source: Prepared by Author from R studio.

The strategic diagram is divided into four quadrants. In the upper right quadrant, the themes are known as motor themes exhibiting high density and centrality, where they signify well-developed, important and fundamental themes in the field of research. The themes in the upper left quadrant are highly developed but isolated themes, indicating well-developed but marginally important for the research area. The lower right quadrant represents basic and transversal themes, while they are important for the research area but still underdeveloped. Lastly, the lower left quadrant is low in density and centrality, reflecting either emerging or declining themes.

The current study presents weather forecasting, feature selection, feature extraction, machine learning, artificial intelligence, and numerical weather prediction as motor themes. Suggesting these themes as well-established, fundamental, and important for this field of research, while these themes are majorly explored by the researchers and presents a scope for future research. Secondly, satellite observations, data mining and remote sensing are niche themes. Presenting them as well-explored and well-established themes, however they do not reflect any importance for the field of research. Thirdly, random forests, solar forecasting, artificial neural network, wind power forecasting and neural network present emerging or declining themes. Lastly, weather prediction and nowcasting are basic and transversal themes representing the important yet underdeveloped themes. However, this quadrant also reflects artificial neural networks as one of the themes, suggesting it as important, underdeveloped and as emerging theme in the field of artificial intelligence and weather forecasting.

Future prospective of climate forecasting with AI

Climate forecasting enabled by AI delivers notable improvements in timeliness, granularity, and accuracy. Large volumes of climatic data may be analyzed by it, leading to more accurate forecasts. High-resolution regional forecasts can be produced by AI algorithms, allowing for tailored predictions for regions that are vulnerable to extreme weather events. Early warning systems for extreme weather events can be made possible by AI-powered forecasting, which can identify early warning indicators of disasters and send out timely alerts. By simulating different scenarios, AI can help with planning for resilience and adaptation to climate change. By enhancing parameterizations and lowering biases, it can be used in conjunction with conventional climate models. AI systems can adjust to real-time data, guaranteeing current and accurate information in a setting that is changing quickly. Decision-makers in a variety of industries can be strengthened by AI-powered climate forecasting, which can also optimize resource allocation, risk management techniques, and climate change resilience. The following perspective of climate forecasting with AI find out in this research.

Research publication. The study provide that limited research on AI climate foresting modeling. There is challenges in accuracy, time and variable in existing model. Future research are more focus on different variables.

Accuracy perspectives. The using AI model, accuracy of forecasting would be increased in term of time, quality and indexed. The real time information will be available with high accuracy. Existing AI model required more explainable like XAI model (Zhou et al., 2023).

Region-wise forecasting. Artificial intelligence can provide region wise information with high resolution equipment and software. The information may be used for planning in different sector like construction, manufacturing, transportation which can reduce overall cost of product (Fu et al., 2023).

Early warning. The AI technique used for early warning information regard to weather forecasting and climate forecasting. This information useful for various stakeholder in the region as well as save life of people.

Supportive to different sector. AI techniques used for information of data for weather forecasting like water storage availability. This information is very helpful for management of resources available with department.

Sustainability. Sustainable environment is challenge in contemporary world. Existing AI based climate forecasting models are predicting climate variable which can be added with sustainable variables.

Discussion

This paper summarizes the research literature of artificial intelligence and weather forecasting from 1984 to 2023 using review of paper and Bibliometric analysis with help of software R studio. This article also highlights the quantitative and intuitive evaluation of academic progress and growth in this field. The research focuses on existing AI role in climate forecasting with help of existing research paper published and With the help of publication trends, the evolution of this field and knowledge structure is demonstrated.

This addresses a core issue in the previous studies, where articles with artificial intelligence and weather forecasting did not reveal the overall changes and bibliometric review. By studying the publication trend of this field, the development and research directions of artificial intelligence and its use in weather forecasting along with its evolution can be tracked.

This paper analyses the overall development of this research theme, providing the expansion of the theme in the upcoming future. Therefore, it also offers guidance for the upcoming researcher and provides in-depth insights into the future growth of this field. Based on bibliometric analysis the results revealed that the first article has been published in 1984 which is four decades old, with respect to that four distinct phases were developed and analyzed. Where the annual growth rate in the publication of literature has been found to be 10.32%. The current phase (2014-2023) witnessed the highest surge in publication and research diversification. One of the most notable findings has been the growth rate of the topic, approximately 88 percent of the total articles are in this phase only. The study has presented the research progress and the most prolific authors, articles and journals. The study has also tried to cover all the aspects related to artificial intelligence and weather forecasting from past research to future research directions.

The analysis was performed to find out the most relevant research and their influence on the topic. Co-word co-occurrence analysis was performed to find out the most common words associated with the theme called clusters (Kamyab et al., 2023). Furthermore, these clusters were analysed to find out the significant association with the theme. After finding the cluster and their correlation, the study took one step further and performed thematic analysis. By utilizing thematic analysis, the study tried to inform the various regions related to the theme and their importance in the upcoming time. The thematic analysis resulted in the identification of a few major themes as artificial neural networks, nowcasting, weather prediction, machine learning, weather forecasting, feature selection, feature extraction, and numerical weather prediction. Highlighting their significance to be explored as future research agenda.

Key findings

The study conducted on the basis of previous literature about AI technology used in Enhancing Climate Forecasting and it future prospective. The study finds that different techniques of AI used in climate furcating which outcome is different based on the model. Bibliometric analysis conducted bases on AI used in climate forecasting. R studio used for analysis of research trends and important literature analysis about AI technology using in current scenario. The finding of study is as under:

As one of the world’s toughest challenges, combating climate change is another area where AI has transformational potential. With the prospective intervention of AI it can be trained to check and measure changes in iceberg melting and mapping deforestation through AI techniques.

Enhancing AI models for climate forecasting

Region-specific forecasting

Collaboration and data sharing

Ethical and responsible AI use

Sustainable practices

Apart from using AI is conserving biodiversity predicting climate change and predicting climatic disasters is strong prospects for AI for future global economies and saving humanitarian disaster by having a proactive approach in disaster management.

Implications

The present study provides several implications for researchers, entrepreneurs, and academicians. Due to its wide range of articles from 1984 to 2023, the study can be utilized in various ways. Also, by knowing the most influential authors, articles and sources of this research domain one can easily identify and associate their work. The study would guide them in future directions by knowing the current existing gap in the literature which could help them to conduct future studies. The study provides an overall view regarding the theme so that the upcoming research can utilize the sources to receive and publish their research in relevant sources. Organizations and entrepreneurs can use the study to predict the upcoming trends in the field and experts to work with them. Technologies powered by Artificial Intelligence techniques, with extreme climate aid tech devices (Dikshit et al., 2024), have advanced dramatically, initiating breakthroughs in additional research sectors. Although a lot of individual Earth System features have been analysed with AI techniques, more general application to comprehend better the complete climate system has not occurred, and the skill to do so may be quite far from the current state of development. At this stage of development, Artificial intelligence (AI) can be used to analyse alert warning mechanism and future predictability numbers with reference to climate foreseen challenges globally. It sensory devices can turn a strong necessary requirements to make the economies prepared for such massive climatic transitions and challenges. AI can aid in understanding and improving existing data and simulations, as it has done in other systems. For instance, Airbus Defence and Space is using TensorFlow, the open-source set of AI tools from Google, to extract information from satellite images and offer valuable insights to customers. The study builds relevant implication and also provides base for future research and practical application.

Research limitations

No research comes without limitations, there are some limitations in this study as well. First and foremost, due to a lack of sources the study utilized only two databases namely Scopus and Web of Science. Future researchers can use other available databases also. The analysis in the study was performed using a single software, researchers may use other or multiple software in their study. The study consists of research articles in the English language only excluding other languages. Future researchers can add different document types to their study, as this research is purely used research articles.