Volume 9, Issue 8, August – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24AUG020

Leveraging Artificial Intelligence to

Address Climate Change
T. Charith Kumar1 U. Sandeep2
Department of Computer Science and Engineering Department of Computer Science and Engineering
Koneru Lakshmaiah Educational Foundation Koneru Lakshmaiah Educational Foundation
Vaddeswaram, A.P, India Vaddeswaram, A.P, India

T. Sushma Nagasri3 P. Sai Kumar4

Department of Computer Science and Engineering Department of Computer Science and Engineering
Koneru Lakshmaiah Educational Foundation Koneru Lakshmaiah Educational Foundation
Vaddeswaram, A.P, India Vaddeswaram, A.P, India

K. Swathi5
Department of Computer Science and Engineering
Koneru Lakshmaiah Educational Foundation
Vaddeswaram, A.P, India

Abstract:- The paper explores how AI-enabled utilizing I. INTRODUCTION

data analytics and machine learning methodologies
enables deeper insights into the intricate patterns and Severe weather patterns, increasing ocean levels, and
behaviors of climate dynamics by analysing amounts of disruptions in nature. It's important to find ways to mitigate
various data, integrating information from various drastic weather fluctuations and employ energy more
origins, like satellite imagery, and the sensory data is efficiently to protect our forests and Generating more
processed to reveal meaningful insights for better polished alternatives using different phrasing to do things, so
understanding and informed actions. These can inform we don't make our home too uncomfortable for everyone.
any policy decisions and facilitate more targeted When we use things like cars, factories, and power plants,
interventions to mitigate the impacts of the climate we often use fossil fuels like gas, coal, and oil. These
conditions. The work discussed here in this research releases a lot of carbon dioxide (CO2) into the air which
provided sources focuses on leveraging artificial forms a layer by trapping the heat and making it warmer.
intelligence (AI) and machine learning (ML) to address Combining all these elements is akin to turning up the
climate change challenges. temperature in our home, and it's becoming a big problem.
The Earth is getting too warm, causing things such as severe
Studies emphasize AI-driven strategies for climate weather, elevated sea levels, and disturbances in natural
change adaptation and including predicting various ecosystems.
changes in the environment, and changes in the weather
patterns. The research highlights the importance of Scientists are trying to understand the Earth's weather
weather conditions, and change in the weather patterns, and climate. Technologies such as artificial intelligence (AI)
and in developing effective AI-powered climate change in can be part of such measures: Assisting in mitigating climate
the adaptation strategies. And accordingly, these studies change, they also aid in addressing various environmental
shows how effectively different AI and ML models like concerns. AI can assist in the identification of areas that are
LSTM, ANN, CNN in improving the climate predictions high at risk of climate change and It predicts the weather
and understanding the weather. AI and ML technologies conditions .AI can predict future climate changes and help
in enhancing the changes in the weather, mitigation. find solutions to problems like pollution or extreme weather
.So, in simple terms, this paper explains how AI and climate
Keywords:- Climate Prediction, Weather Events, Climate researchers team up. These appliance plays a key role in
Change, Weather Forecasting, Machine Learning climate-related tasks. AI's capability to swiftly analyse vast
techniques, Weather Patterns, AI-driven Strategies, Climate amounts of climate data aids in identifying trends and
Data Analysis. comprehending how our climate is evolving Similar to a
weather forecaster predicting future conditions, AI can
forecast climate changes, enabling us to prepare for events
like storms or heat waves.

IJISRT24AUG020 www.ijisrt.com 958

Volume 9, Issue 8, August – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24AUG020

 Uncertainty in Predictions:
AI models may struggle with uncertainties inherent in
climate systems. Climate is influenced by numerous
variables, and predicting complex interactions accurately
can be challenging, leading to uncertainties in long-term
forecasts.

 Complex Interactions:
Capturing the intricate interplay between variables like
ocean currents, atmospheric circulation, and human activity
within climate models is a major challenge.

 Long-Term Forecasts :
Predicting climate changes decades into the future
Fig 1 The Schematic Representation of a Machine Learning becomes increasingly unreliable due to the accumulation of
Process for Climate Prediction. uncertainties over time.
Furthermore, AI functions as a wise advisor by offering
These challenges highlight the need for:
intelligent suggestions on pollution reduction or adapting to
changes, guiding the formulation of effective climate
 Improved Data Collection:
policies. As references from the several research paper and Expanding global climate collection efforts,
article we came to know that there are some limitations
particularly in underrepresented regions, is crucial.
faced during the process.
 Data Bias Detection and Correction:
II. CHALLENGES OCCURRED Techniques to identify and mitigate biases in climate
datasets are indispensable for ensuring the reliability of AI
 Data Dependency:
models.
AI heavily relies on the data it's trained on. If the data
is incomplete or biased, the AI's predictions and suggestions  More Efficient AI Algorithms:
may be skewed or inaccurate, potentially leading to Developing more efficient AI algorithms that require
misguided climate policies. less computational power can address resource limitations.
 Incompleteness:  Uncertainty Quantification:
Climate data collection can be sparse, especially in the AI models should be able to quantify the level of
developing regions. Missing data can lead to biases in the uncertainty associated with their predictions, allowing for a
model and inaccurate predictions. more understanding of potential climate futures.
 Bias: By addressing these challenges, AI can become a
If the training data is biased towards certain weather powerful tool for Enhancing climate predictions to inform
patterns or regions, the AI model might lack the capability to efficient mitigation and adaptation strategies.
reliably forecast future changes in other areas or under
unforeseen circumstances. III. LITERATURE REVIEW
 Resource Intensiveness: [1] AI-enable Adjusting to a warming world by using
AI solutions, particularly deep learning models, can be less water, planting smart, building strong, planning for
computationally expensive and require substantial resources. emergencies, and working together. involve expert oversight
This might present obstacles to widespread acceptance and interpretation to ensure effective and ethical use.
especially in regions with limited technological Projects Deltares in the Netherlands employs AI to model
infrastructure. sea level increase impacts on coastal infrastructure,
combining leveraging machine learning with large datasets
 Accessibility: to forecast future climate changes Ethical considerations,
Resource-limited regions may not have the transparency, inclusivity, and bias mitigation are crucial in
infrastructure to run these models, hindering widespread developing AI- powered different areas can adapt to climate
adoption of AI-based climate prediction tools. change by using smarter protections of all. Jain, H.,
Dhupper, R., Shrivastava, A. et al. AI- enabled strategies for
 Sustainability: climate change adaptation: protecting communities,
The high energy consumption of training this models infrastructure, and businesses from the impacts of climate
raise apprehensions regarding their environmental impact. change. Compu Train Sci. 3, 25 (2023).

IJISRT24AUG020 www.ijisrt.com 959

Volume 9, Issue 8, August – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24AUG020

[2] The paper discusses the use of LSTM for climate forecast. Felsche, E. and Ludwig, R.: Powerful computer
prediction, weather forecasting, Additionally, they utilize AI models analyze massive climate data sets (large ensemble
for other time series forecasting tasks, highlighting its simulations). This analysis helps predict future droughts,
effectiveness in capturing long-term dependencies in allowing for smarter adaptation strategies. Nat. Hazards
sequential data. explores the accuracy and Algorithms Earth Syst. Sci., 21, 3679–3691.
employed for analyzing time series datasets forecasting of
environmental variables. The study focuses on forecasting
snow cover, temperature, and normalized difference
vegetation index (NDVI) in the Himachal Pradesh region
using ANN and LSTM(Long Short Term Memory) models.
ANN, specifically a Multi-layer Perceptron (MLP), is
employed to learn complex patterns in structured data, while
LSTM, a type of Recurrent Neural Network (RNN), is
utilized for its ability to capture long-term dependencies in
time series data. The research compares the effectiveness of
ANN and LSTM models. Haq, M.A., Ahmed, A., Khan, I. et
al. Analysis of environmental factors using AI and ML
methods. Sci Rep 12, 13267 (2022).

[3] AI and ML techniques alongside traditional

physical models to enhance the accuracy of seasonal
forecasts, particularly for critical climate parameters like
monsoon rainfall. By integrating historical data and key
climate drivers like ENSO and IOD, Climate models can
guide smarter adaptation: from future-proof farms to flood-
ready cities and personal eco-actions, predictions compared
to conventional methods. The study showcases the Fig 2 First four leading eigenfunctions the level of the sea
effectiveness of Long Short-Term Memory (LSTM) models and pressure in CanESM2. Percentage of variance the mode
in predicting All India Monsoon Rainfall (AISMR). The is displayed a top the panels. Cal orthogonal functions
algorithms include linear regression, ARIMA, SARIMA, (EOFs)(Dawson,2016). The leading modes of the PCA
Long Short-Term Memory, SVR, XGBoost, and CNN corresponding to NAO, SCA, EA and EAWR derived from
Models.The research compares performance of these models the CanEsm2 dataset in Fig 2
using different datasets: historical AISMR alone, historical
AISMR with the Niño3.4 index, and historical AISMR with [5] Decision tress and neural networks the algorithms
the Niño3.4 index and categorical Indian Ocean Dipole optimize non-linear regression, where deep leaning using
(IOD) data. The models are well trained on datasets dimensional models. The methods accuracy is dependent on
incorporating additional climate drivers like Niño3.4 and some characteristics and extensive iteration to determine the
IOD outperform those trained solely on historical AISMR optimal architecture. ML architectures like random forests
data. Specifically, the LSTM model stands out as the best (RF), gradient- boosted decision trees, and Support vector
performing model, demonstrating higher accuracy and lower machines are commonly utilized, each carrying its own
error rates compared to other models. Narang, U., Juneja, advantages and disadvantages. Recent studies focus on ML
K., Upadhyaya, P. et al. Artificial intelligence predicts for parametrization and emulation of sub grid-scale
normal rainfall in 2023. Sci Rep 14, 1495 (2024). processes, such as radiative transfer, with examples like
Chevallier et al. (1998) training NNs to represent radiative
[4] AI methods like Shapley Additive explanations transfer budgets efficiently. Additionally, researchers are
(SHAP) crack the code on future risks, guiding smarter exploring uncertainty quantification and capturing extremes
ways to adapt and emphasizes the influence of key variables in ML models for weather and climate applications. The
such as the North Atlantic Oscillation index on accurate essence of machine learning lies in the automation involved
prediction. Additionally, it discusses the relevance of in training models, where complex Models of the Earth’s
performance metrics like accuracy, F1 score, and Heidke require a blend of architectures and algorithms to emulate
skill score (HSS) in evaluating model performance. . The components effectively. de Burgh-Day, C. O. and Leeuwen
accuracy is evaluated using two key metrics: accuracy and burg, T.: Machine Learning for numerical weather and
F1 score. The F1 score, especially useful for imbalanced climate modelling: a review, EGU sphere
class distributions, was found to be 0.26 for Lisbon and 0.24
for Munich due to the class imbalance within the dataset. [8] The accuracy and algorithms used in predicting
The investigation necessitated accuracy assessment for weather patterns. Powerful image-like analysis (CNNs)
every class. be at least 50%. Additionally, the best- excels at recognizing and predicting repeating weather
performing models were assessed using the Heidke skill patterns over 5 days. With enough data (1000+ samples),
score (HSS), which ranges from -∞ to 1, indicates zero accuracy tops 90% for both tasks. The accuracy of
below the value that a random forecast performs better than prediction weakly scales with The lead days number exhibits
the trained model, while an HSS of 1 signifies a perfect non-linear variation to the size of the training set. Testing

IJISRT24AUG020 www.ijisrt.com 960

Volume 9, Issue 8, August – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24AUG020

different brain-inspired computer models (CNNs) for precipitation, temperature, pressure, and radiation) and
weather pattern recognition. A more complex model methods (Deep Learning, Random Forest, XGBoost ,ANN
(CNN4) performed better than a simpler one (CNN2). and SVM) in these topics. Meteorological fields in NWP
CNN4 achieved over 93% accuracy in both summer and studies, such as wind, precipitation, temperature, air
winter re-identification tasks. This means the model excelled pressure, and radiation, and the most used methods, involves
at recognizing repeating weather patterns. With an accuracy ANN and Deep machine learning methods used in weather
margin of ±0.2%, the results were highly consistent. This forecasting and climate analysis, exploring its applications
paves the way for more precise weather forecasting. and future possibilities. Chris Huntingford1, Elizabeth S
Powerful image-like analysis (CNNs) excels at recognizing Jeffers2, Michael BBonsall2, HannahM Christensen3 and
and predicting repeating weather patterns over 5 days, With Hui Yang1,5.
enough data (1000+ samples), accuracy tops 90% for both
tasks. Chattopadhyay, A., Hassanzadeh, P. & Pasha, S. [7] evaluates the accuracy and the models predicts the
Predicting clustered weather patterns: A test case for microclimate variations. Northern Italian researchers used
applications of convolutional neural networks to spatio- advanced neural networks to predict local temperature and
temporal climate data. Sci Rep 10, 1317 (2020). humidity changes. The models underwent training using
global climate data (ERA5).and local weather station
[10] The future of machine learning (ML) and artificial measurements (ARPA). Researchers evaluated model
intelligence (AI) to aid the research of change of climate. accuracy with metrics like mean error (MAE) - but for
Machine learning unlocks hidden insights from climate data temperature, a "goodness of fit" score (R²) is more
and simulations. This empowers smarter predictions and important. This is because temperature naturally fluctuates
adaptation strategies and to reduce inter-ESM uncertainty. daily, so capturing the overall trend matters more than
ML and AI can be used to understand and capitalize on pinpointing exact values. The research also addressed the
existing data and simulations, and to reduce inter-ESM black-box nature of neural networks by proposing a method
uncertainty. Among the techniques employed are Artificial to interpret the significance of input variables on the output.
Neural Networks (ANN),Convolutional Neural Local weather data yielded the most accurate microclimate
Network(CNN), and others Gaussian Process (GP) predictions, but global data (ERA5) served as a good
Regression, and others. AI smarts analyse climate data, alternative when local data was missing. This means we can
pinpointing risks and guiding better adaptation. Machine still make informed predictions even in areas with limited
learning muscles power climate solutions, from smarter weather stations. This research underscores the potential for
farms to flood-ready cities. Published 22 November 2019 • advanced neural networks in predicting microclimate
© 2019 The Author(s). Published by IOP Publishing Ltd. variations with high accuracy. By combing global and local
This study analyses 500 recent scientific articles exploring data sources and employing interpretable models,
how machine learning is revolutionizing climate and researchers can provide valuable insights into local climate
weather prediction. The article highlights the most common dynamics, enabling better-informed decision-making in
topics of interest in the abstracts, including photovoltaic and various sectors such as agriculture, urban planning, and
wind energy, atmospheric physics and processes, disater management. ] Zanchi, M., Zapperi, S. & La Porta,
parametrizations, extreme events, and climate change. It also C.A.M.(2023).
identifies the most examined meteorological fields (wind,

Fig 3 The Figure Shows the Difference between the ERA5 Database and the ARPA Station Recordings [7] Marco Zanchi
Article Number: 21062 (2023)

IJISRT24AUG020 www.ijisrt.com 961

Volume 9, Issue 8, August – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24AUG020

[6] approach using a CNN Model to analyze extreme IV. METHODOLOGY

precipitation sensitivity to climate change. By training the
CNN on precipitation data from 10,000 stations, This From table 1. It table show the results of Weather
research pinpoints how heavily extreme rainfall will be prediction models using the Long Short-Term
affected by climate change across vast regions. It covers Memory(LSTM), Convolutional Neural Network(CNN),
North America, Europe, Australia, and New Zealand, Feedforward Neural Network(FNN) models. The table
revealing detailed variations in these impacts. The research compares the performance of these three models based on
aims to address the challenges posed by limited and four metrics: accuracy, precision, recall, and f1 score. Based
heterogeneous observational records in studying extreme on these we can consider which model is more accurate and
precipitation changes with climate warming. Researchers can be used for future enhancements.
trained machine learning models on a massive dataset
(10,000 stations) to predict extreme rainfall events. The  Accuracy:
models analyze how these events will be impacted by rising Is the overall proportion of correct predictions made by
global temperatures across different regions. Modeling to the model. In the table, all three models have an accuracy of
analyze extreme precipitation sensitivity to climate change. approximately 97%.
A special AI regional vulnerabilities to future downpours
Another example is a DL model that was used to predict  Precision:
extreme precipitation events in the United State. Trained on Is the ration of true positive predictions to the total
past weather, this powerful model predicts extreme rainfall number of positive predictions. In the table, all three models
events with impressive accuracy. This allows for early have a precision of 98%.
warnings and better preparedness for downpours.
Model(CNN) analyzed rainfall data from 4 continents to  Recall:
map how extreme weather will vary with raising Is the ratio of true positive predictions to the actual
temperatures. This helps us understand the DL models in positive cases. In the table, all three models have a recall of
climate change predictions are more accurate to predict the 97%.
climate events. This research not only addresses challenges
related to limited observational records but also facilitates  F1 Score:
early warnings and enhances preparedness for extreme Is a harmonic mean between precision and recall. It
weather events, contributing to improved resilience in the provides a more balanced view of a model’s performance
face of climate change. Bodeker, G.E.(2023). than just using precision or recall alone. In the table, LSTM
and CNN models have an F1 score of 98%, while the FNN
model has an F1 score of 97%.

Table 1 A Concise Comparison of the Performance Metrics Across Different AI Algorithms for the Weather Prediction on a
Dataset
Metrics Long Short-Term Memory Convolutional Neural Network Feedforward Neural Network
(LSTM) (CNN) (FNN)
Accuracy 98.39% 97.03% 98.19%
Recall 98.14% 97.21% 96.62%
Precision 98.46% 98.74% 96.29%
F1 Score 98.80% 97.46% 97.44%

From table 1. The LSTM and CNN models appear to

perform almost identically, with a slight edge to the CNN
model in terms of F1 score. The FNN model performs
slightly lower than the other two models in all metrics. It is
important to note that the performance of a model can vary
depending on the specific weather prediction task and the
dataset used to train the model.

Fig 4 Performance Analysis of LSTM, CNN and FNN

Algorithms for Weather Prediction.

IJISRT24AUG020 www.ijisrt.com 962

Volume 9, Issue 8, August – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24AUG020

The metrics collectively offer a comprehensive [5]. de Burgh-Day, C. O. and Leeuwenburg, T.: Machine
understanding of the model outcomes. The Long Short-Term Learning for numerical weather and climate
Memory(LSTM) demonstrates strong overall accuracy, modelling: a review, EGUsphere [preprint],
precision, recall and an effective handling of positive and https://doi.org/10.5194/egusphere-2023350, 2023
negative instances. Evaluation metrics are crucial for [6]. Bird, L.J., Bodeker, G.E. & Clem, K.R.
assessing the suitability of a model for specific tasks. Sensitivity of extreme precipitation to climate change
inferred using artificial intelligence shows high
V. CONCLUSION spatial variability. Commun Earth Environ 4, 469
(2023). https://doi.org/10.1038/s43247-023-01142-4
The research discussed in the article focuses on [7]. Zanchi, M., Zapperi, S. & La Porta, C.A.M.
leveraging artificial intelligence (AI) to address climate Harnessing deep learning to forecast local
change challenges. Various studies highlight the use of AI- microclimate using global climate data. Sci Rep 13,
enabled strategies for climate change adaptation, including 21062 (2023). https://doi.org/10.1038/s41598-
predicting changes in the environment and weather patterns. 02348028-1
AI and ML technologies play a crucial role in enhancing [8]. Chattopadhyay, A., Hassanzadeh, P. & Pasha, S.
climate predictions, understanding weather changes, and Predicting clustered weather patterns: A test case for
guiding effective climate policies. However, the research applications of convolutional neural networks to
also points out some limitations such as data dependency, spatiotemporal climate data. Sci Rep 10, 1317 (2020).
resource intensiveness, and uncertainties in predictions. and https://doi.org/10.1038/s41598-020-57897-9
develop effective adaptation strategies, but it also [9]. Yunjie Liu, Evan Racah, Prabhat, Joaquin Correa.
underscores the importance of addressing challenges like Application of Deep Convolutional Neural Networks
data quality, resource requirements, and prediction for Detecting Extreme Weather in Climate
uncertainties to ensure the effectiveness and reliability of Datasets.Sci Rep Wed, 4 May 2016 06:38:19 UTC
AI-powered climate interventions. https://doi.org/10.48550/arXiv.1605.01156
[10]. Chris Huntingford1, Elizabeth S Jeffers2, Michael B
Form figure 4. for further future enhancements we can Bonsall2, Hannah M Christensen3 and Hui Yang1,5 .
consider Long Short-Term Memory(LSTM). By addressing Machine learning and artificial intelligence to aid
accuracy improvement, precision and recall, f1 score climate change . Published 22 November 2019 • ©
enhancement, model comparison and selection and 2019 The Author(s). Published by IOP Publishing Ltd
potentially exploring advanced techniques in model Environmental Research Letters, Volume 14, Number
optimization, data preprocessing, and the weather prediction 12 Citation Chris Huntingford et al 2019 Environ.
models utilizing LSTM can be refined to provide more Res. Lett. 14 124007DOI 10.1088/1748-9326/ab4e55
accurate, reliable, and efficient forecasts for climate-related [11]. Bochenek, B.; Ustrnul, Z. Machine Learning in
tasks. Weather Prediction and Climate Analyses—
Applications and Perspectives. Atmosphere 2022, 13,
REFERENCES 180. https://doi.org/10.3390/atmos13020180
[12]. Chen, L., Chen, Z., Zhang, Y. et al. Artificial
[1]. Jain, H., Dhupper, R., Shrivastava, A. et al. AI- intelligence-based solutions for climate change: a
enabled strategies for climate change adaptation: review. Environ Chem Lett 21, 2525–2557 (2023).
protecting communities, infrastructure, and [13]. Rutenberg, I., Gwagwa, A., Omino, M. (2021). Use
businesses from the impacts of climate change. and Impact of Artificial Intelligence on Climate
Comput.Urban Sci. 3, 25 (2023). https://doi.org/ Change Adaptation in Africa. In: Oguge, N., Ayal, D.,
10.1007/s43762-02300100-2 Adeleke, L., da Silva, I. (eds) African Handbook of
[2]. Haq, M.A., Ahmed, A., Khan, I. et al. Analysis of Climate Change Adaptation. Springer, Cham.
environmental factors using AI and ML methods. Sci https://doi.org/10.1007/978-3-030-45106-6_80
Rep 12, 13267 (2022). https://doi.org/10.1038/ [14]. Cowls, J., Tsamados, A., Taddeo, M. et al. The AI
s41598-02216665-7 gambit: leveraging artificial intelligence to combat
[3]. Narang, U., Juneja, K., Upadhyaya, P. et al. Artificial climate change— opportunities, challenges, and
intelligence predicts normal summer monsoon recommendations. AI & Soc 38, 283–307 (2023).
rainfall for India in 2023. Sci Rep 14, 1495 (2024). https://doi.org/10.1007/s00146021-01294-x
https://doi.org/10.1038/s41598-02344284-3 [15]. Coeckelbergh, M. AI for climate: freedom, justice,
[4]. Felsche, E. and Ludwig, R.: Applying machine and other ethical and political challenges. AI Ethics 1,
learning for drought prediction in a perfect model 67–72 (2021). https://doi.org/10.1007/s43681-
framework using data from a large ensemble of 02000007-2
climate simulations, Nat. Hazards Earth Syst. Sci.,
21, 3679–3691, https://doi.org/10.5194/nhess-21-
36792021, 2021.

IJISRT24AUG020 www.ijisrt.com 963