AI-Enabled decision support systems for agrometeorological forecasting and climate change mitigation

Introduction

Agriculture is a leading climate sensitive sector and weather variability and extreme high and low climate events can detrimentally impact crop production and water resources, and in turn food security. By aiding farmers and policymakers in making informed decisions when faced with the risks of the unpredictable weather patterns and climate change, agrometeorological forecasting acts as an extremely important tool. Historical climate data and empirical models often used in traditional forecasting methods are no guarantee that it will full capture the complexities of evolving climate dynamics [1] . In this regard, Artificial Intelligence (AI) supported Decision Support Systems (DSSs) have evolved into very powerful tools for improving the accuracy of forecasting and the optimal adaptation strategies in agriculture in the face of climate change. DSS using artificial intelligence (AI) #driven incorporation of machine learning (ML), deep learning (DL), and big data analytics to work on high volumes of meteorological, soil and crop oriented data [2] . These systems allow for predictive modeling of weather patterns; early warning for climate oddities; real time decision making for increased agricultural production and sustainability. Traditionally, complex climate data have been analyzed using simple statistical methods, however, AI algorithms, namely neural networks, support vector machines and ensemble models have shown to outperform these methods [3] . In addition, AI enabled DSS is also suitable to support precision farming by tailoring irrigation schedules, pest controls, different crop selection to avoid wastage and environmental harm. Additionally, AI enabled systems help to reduce climate change by monitoring carbon footprint, monitoring of greenhouse gas emissions, and promoting sustainable land management.

By integrating AI to Internet of Things (IoT) devices, satellite imagery, and remote sensing technologies, real-time data collection is amplified for improved data collection on forecasting procedures. But issues like data quality, model interpretability, and the ones related to ethical use of AI in agriculture all need to be addressed. In this research, we investigate the feasibility of the use of AI based DSS to enhance agrometeorological forecast in terms of promoting climate resilience and sustainable agricultural practice. AI bridges the technology and environmental management gap where solutions are available to foster a climate resilient future in agriculture.

Related Works

In recent years, agrometeorological forecasting and adaptation to climate change have received great attention and numerous studies have been exploring the use of AI based approaches to improve the predictive accuracy and agro-weather decision making. In searching for better climate change, agricultural risk, and sustainable farming, various machine learning (ML) and deep learning (DL) models have been applied.

AI-Enabled Climate Forecasting for Agriculture

In recent years, the potential of AI for improved climate forecasting for agriculture has been proven by several studies. Hamam et al. [15] looked into farmers' perceptions of climate change and farmers' mitigation practices in Hungary, and the need for an integration of AI based forecasting tools to support farm decision making. The work they reported showed how AI can help make more efficient use of the available resources, cut down on climate losses, and boost productivity. This was also done in a study of Kennedy et al. [18] who addressed precipitation trends and snow cover in the Andean region, presenting their implication to hydrological resources. The study supports the utility of AI models in analyzing large scale climate patterns as an aid to water management in agriculture. Apart from this, Maake et al. [22] designed the Umlindi Newsletter which is a climate information dissemination system that allows for the real time updates on the weather conditions, disasters and the agricultural production in South Africa. To mitigate the risk and enable adaptive decision making, it highlights the value of using AI in climate communication.

Machine Learning for Crop Yield Prediction and Climate

Adaptation ML has been used to predict crop yield in several studies and to assess climate suitability for different crops. Jayasinghe and Kumar [17] conducted a systematic review on the effect of the climate change on the tea production projects. They emphasized that AI models can be applied for the climate suitability assessments. It highlighted that invoking ML based climate prediction with crop needs is likely to enhance crop resilience and sustainability. In their work, Maryelle et al. [24] used machine learning to develop models of the impacts of extreme climatic on dwarf green coconut trees in the Eastern Amazon. However, their findings found that AI models could accurately predict the productivity decline of climate variations and could contribute to the formulation of adaptive farming strategies. Like Mamassi et al. [23] , they also developed a genotype × environment × management (G × E × M) interaction model at the cropping systems stage for rainfed wheat cropping systems in Morocco. AI driven predictions of wheat productivity under variable climatic conditions were incorporated for the purpose of sustainable intensification. Liu et al. [21] also used AI for drought susceptibility assessment using modeling and projections of the climate to enhance land use planning and water resource management.

Remote Sensing and AI for Agricultural Disaster Prevention

Remote sensing with AI has been found to be an effective means of preventing agricultural disaster and its assessment. In Taiwan, Ming-Hwi Yao et al. [25] used AI derived insights for agricultural disaster prevention system. ML models can predict extreme weather events, but their impact on agricultural productivity, which was demonstrated by their research. Based on this, Li et al. [20] suggested some dynamic monitoring mechanisms for low temperature disasters on winter wheat by integrating AI based mitigation mechanisms and environmental coercion factors. AI has been proven to help identify early warning signals and enhance climate resilience in agriculture, according to their study. They [16] Isilda et al. explored effects of tree shade on vineyard microclimate and on vineyard grape production, introducing a new approach to modeling sun radiation related to climate change. Their work showed how models of microclimate using AI could aid in the optimization of vineyard management in changing environments. As is also done by Krasovska et al. [19] , they assessed the risks related to drained lands with AI based risk assessment models aiming to assist in making water management and agricultural planning decisions.

Advancements in Climate Adaptation and Mitigation Services

There exist many studies which point out the necessity of having AI powered climate adaptation and mitigation services in agriculture. Moelesti and Tsubo [26] suggested that it was also necessary to develop usable climate services that are appropriate for the agricultural sector. AI integrated climate information systems for boosting farmers decision making capacity and adaptation strategies to climate change were the research. Carried further with reference to Li et al. [20] who further showed how AI can generate robust climate adaptation frameworks by leveraging real time weather data, predictive analytics. The ways in which AI is being applied into climate change mitigation are not limited to forecasts for further mitigation. According to Mamassi et al. [23] , as well as Liu et al. [21] , AI will be used to optimize land use, water resource allocation, and soil health management. Based upon their studies, they propose that AI based decision support systems can substantially enhance agricultural sustainability through its capacity to make resilience to climate variability.

Methods and Materials Data Collection and Processing

The effectiveness of AI enabled Decision Support Systems (DSS) used in agrometeorological forecasting depends on availability of quality and diversity of data. To do so, this research relies on a combination of historical meteorological data, satellite imagery, soil and crop yield data [4] . Publicly available repositories such as NASA Earth Observation, FAO Climate Data, and the rest of regional agricultural monitoring centers are used as data sources. The variables in the dataset are temperature, humidity, precipitation, soil moisture levels, wind speed, radiation indices and so on. To achieve data consistency and improve model accuracy, preprocessing techniques, data normalization, handling missing values and feature engineering are applied to the given sentence [5] . Assuming the input to machine learning (ML) models is in fact structured, it is first reduced in dimensionality with methods like Principal Component Analysis (PCA) to optimally remove the dimensionality without missing the key patterns in the data [6] . The dataset is split into training (70%), validation (15%) and testing (15%) sets for evaluation of the goodness of the AI driven forecasting models. The statistical metrics that are used to measure model performance are Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and R-squared (R²) values [7] .

Selected AI Algorithms for Agrometeorological Forecasting

In the implementation part of this thesis, four ALs, namely, Long Short Term Memory (LSTM), Random Forest (RF), Support Vector Regression (SVR), and Convolutional Neural Networks (CNNs), are employed. The detailed descriptions of each algorithm are given below.

Long Short-Term Memory (LSTM)

LSTM is a recurrent neural network with special purpose for handling a sequential data. In climate prediction, it provides a very good measure of long term dependencies on time series forecasting, and is very effective at that. Here, LSTM uses memory cells that modulates the information flow by means of three gates [8] .  Forget Gate: Choose which in the past information to wipe out.  Input Gate: Updates the memory with new information.  Output Gate: Determines the amount of information transferred from one time step to the next time step.

LSTM has been extensively applied in agrometeorology for forecasting temperature trends, precipitation behavior, and probability of drought. It learns from historical climate variability to predict the future, enhancing the quality of weather forecasts [9] . The structure includes several LSTM layers with dense layers at the end to transform features into output variables.

Random Forest (RF)

Random Forest is a method of ensemble learning that builds several decision trees while training and averages their predictions. It improves prediction accuracy and avoids overfitting by incorporating randomness in feature selection and data sampling [10] . For agrometeorological purposes, RF is employed to classify climate zones, forecast soil moisture content, and estimate crop yield. The algorithm is advantageous due to its capability to process large datasets with high-dimensional features while offering strong and interpretable outcomes [11] .

"Initialize number of trees (N)

For each tree in N: Select random subset of training data Construct decision tree using selected data Perform feature selection and split nodes based on information gain Train tree until stopping criteria are met End loop Aggregate predictions from all trees and compute final forecast" Support Vector Regression (SVR) Support Vector Regression (SVR) is a robust regression method based on Support Vector Machines (SVM). It is intended to identify a function that optimally fits the data within a specified margin, and hence it is applicable for accurate climate forecasting tasks like temperature and rainfall prediction [12] . SVR operates by transforming input data into a highdimensional space via a kernel function (linear, polynomial, or radial basis function). The algorithm subsequently determines a hyperplane that optimizes the error while ensuring that most of the predictions fall in the tolerance margin. SVR finds special application in agrometeorology because it can deal with nonlinear relationships between climate factors and growth factors of crops.

"Initialize SVR model with selected kernel function

Define margin of tolerance (ε) For each data point in training set: Map input to high-dimensional space using kernel function Optimize hyperplane to minimize error while staying within margin ε End loop Predict values for test data and evaluate model performance"

Convolutional Neural Networks (CNN)

Convolutional Neural Networks (CNN) are deep learning networks mostly applied to image processing. CNNs are utilized in agrometeorology to analyze satellite images and weather charts to identify climate patterns, soil health fluctuations, and vegetation cover changes [13] . A CNN is composed of convolutional layers that learn spatial features from input images, pooling layers that down-sample, and fully connected layers that make predictions. The most significant strength of CNNs in agricultural forecasting is their capacity to automatically learn intricate spatial dependencies from multispectral satellite imagery without the need for extensive feature engineering.

Experiments: Experimental Setup

The experiments were carried out on a dataset of agrometeorological parameters gathered from various sources, such as past weather records, satellite images, and IoT-based real-time sensors. The dataset contained parameters like temperature, humidity, rainfall, wind speed, soil moisture, and radiation index [14] . The research aimed to assess the performance of four AI models: "Long Short-Term Memory (LSTM), Random Forest (RF), Support Vector Regression (SVR), and Convolutional Neural Networks (CNN)". The dataset was normalized, missing values were imputed, and feature selection was done. 70-15-15 was split for training, validation, and testing. The models were separately trained for each climate parameter and compared.

Results and Analysis: AI Model Forecasting Accuracy

The initial experiment evaluated how well AI models can predict temperature from past data. The outcomes are presented in Table 1, which includes the RMSE, MAE, and R² scores of each model. it is sensitive to high-dimensional variability in weather data [27] . Multi-Variable Forecasting Performance: In the second experiment, the models were validated for predicting precipitation, humidity, and soil moisture. The results are given in Table 2. • Random Forest recorded the highest RMSE and was therefore less efficient for accurate climate prediction.

Model Training Time and Efficiency

Training time is an important consideration when applying AI models to real-time decision support systems. Table 3 shows the computational efficiency of every model. • Prediction time was shortest for Random Forest and SVR, and hence they are better suited for rapid decisionmaking [28] .

Impact of Data Volume on Model Performance

To determine the impact of dataset size on model accuracy, the models were trained using varying data sizes and their corresponding RMSE values noted. The outcome is presented below in Table 4.  Random Forest performed poorly with large datasets, showing its inability to process high-dimensional agricultural data [29] .

Generalization Ability of AI Models

To test how good the models can generalize over regions of varying climate, the dataset was evaluated against multiple site-based datasets. Summary results are tabulated in Table 5. reflecting sensitivity to regional changes [30] .

Conclusion

This research has investigated the merging of agrometeorological forecasting and climate change mitigation through AI enabled decision support systems of agromeo treatment prognosis. Inhibiting the adverse impacts of climate variance and optimizing resource management, the application of the machine learning (ML) and deep learning (DL) models has achieved benefits on increasing the prediction accuracy of climate. Each AI algorithm was studied including Random Forest, Long Short Term Memory (LSTM), Support Vector Machines (SVM) and Convolutional Neural Networks (CNN), which played a role in the prediction of climate and agricultural risk. Finally, these models helped with a better understanding of the patterns of precipitation, temperature fluctuation, susceptibility to drought, and helped with a more informed decision making in

References

1. Aschale TM, Cancelliere A, Palazzolo N, Buonacera G, Peres DJ. Analysis of the spatiotemporal trends of standardized drought indices in Sicily using ERA5-Land reanalysis data (1950-2023). Water. 2024;16(18):2593.
2. Bouri M, Arslan KS, Şahin F. Climate-smart pest management in sustainable agriculture: Promises and challenges. Sustainability. 2023;15(5):4592.
3. Campillay-Llanos W, Ortega-Farías S, González-Colville P, Díaz GA, López-Flores MM, López-Olivari R. Modeling the effects of extreme temperatures on the infection rate of Botrytis cinerea using historical climate data (1951-2023) of Central Chile. Agronomy. 2025;15(3):608.
4. Cappucci S, Nappi S, Cappelli A. Green public areas and urban open spaces management: New GreenCAL tool algorithms and circular economy implications. Land. 2022;11(6):886.
5. Charalampopoulos I, Droulia F. A pathway towards climate services for the agricultural sector. Climate. 2024;12(2):18.
6. Cruz-Baltuano A, Huarahuara-Toma R, Silva-Borda A, Chucuya S, Franco-León P, Huayna G, et al. Assessment of observed and projected extreme droughts in Perú-Case study: Candarave, Tacna. Atmosphere. 2025;16(1):18.
7. Dhanya P, Geethalakshmi V, Ramanathan S, Senthilraja K, Sreeraj P, Pradipa C, Bhuvaneshwari K, Vengateswari M, Dheebakaran G, Kokilavani S, et al. Impacts and climate change adaptation of agrometeorological services among the maize farmers of West Tamil Nadu. AgriEngineering. 2022;4(4):1030.
8. Du S, Xiong W. Weather extremes shock maize production: Current approaches and future research directions in Africa. Plants. 2024;13(12):1585.
9. Faraz M, Mereu V, Spano D, Trabucco A, Marras S, Chami DE. A systematic review of analytical and modelling tools to assess climate change impacts and adaptation on coffee agrosystems. Sustainability. 2023;15(19):14582.
10. Farmanov T, Umarov S, Hilorme T, Khasanov B, Durmanov A. The impact of energy aspects on the climatic indicators of agricultural products. International Journal of Energy, Environment and Economics. 2022;30(2):187-209.
11. Fonseca A, Cruz J, Fraga H, Andrade C, Valente J, Alves F, Neto AC, Flores R, Santos JA. Vineyard microclimatic zoning as a tool to promote sustainable viticulture under climate change. Sustainability. 2024;16(8):3477.
12. Gouveia C, Justino F, Gurjao C, Zita L, Alonso C. Revisiting climate-related agricultural losses across South America and their future perspectives. Atmosphere. 2023;14(8):1303.
13. Grandis A, Fortirer JS, Navarro BV, de Oliveira LP, Buckeridge MS. Biotechnologies to improve sugarcane productivity in a climate change scenario. Bioenergy Research. 2024;17(1):1-26.
14. Gurdeep SM, Kaur M, Kaushik P. Impact of climate change on agriculture and its mitigation strategies: A review. Sustainability. 2021;13(3):1318.
15. Hamam M, Raimondo M, Spina D, Király G, Di Vita G, D'Amico M, et al. Climate change perception and innovative mitigation practices adopted by Hungarian farms. AGRIS On-line Papers in Economics and Informatics. 2023;15(3):57-72.
16. Isilda CM, Santos M, Bugalho L, Mário GP. The effects of tree shade on vineyard microclimate and grape production: A novel approach to sun radiation modelling as a response to climate change. Land. 2024;13(11):1970.
17. Jayasinghe SL, Kumar L. Potential impact of the current and future climate on the yield, quality, and climate suitability for tea (Camellia sinensis): A systematic review. Agronomy. 2021;11(4):619.
18. Kennedy DA, SM, Rafael CD, Ana Pâmela Tavares Mendonça. Precipitation trends and Andean snow cover: Climate interactions and hydrological impacts in the Acre River Basin (1982-2023). Atmosphere. 2025;16(3):249.
19. Krasovska Y, Lesniak O, Podlevska O, Podlevskyi A, Kostrychenko V. Consideration of risks in assessing the potential of drained lands: National and international aspects. IOP Conference Series: Earth and Environmental Science. 2024;1429(1):012011.
20. Li M, Huo Z, Mi Q, Zhang L, Yang J, Zhang F, et al. A dynamic monitoring framework for spring lowtemperature disasters affecting winter wheat: Exploring environmental coercion and mitigation mechanisms. Agronomy. 2025;15(2):337.
21. Liu J, Li M, Li R, Masoud JS, Ren Y, Silakhori E. Comprehensive assessment of drought susceptibility using predictive modeling, climate change projections, and land use dynamics for sustainable management. Land. 2025;14(2):337.
22. Maake R, Malherbe J, Masupha T, Chirima G, Beukes P, Roffe S, et al. The Umlindi Newsletter: Disseminating climate-related information on the management of natural disaster and agricultural production in South Africa. Climate. 2023;11(12):239.
23. Mamassi A, Balaghi R, Devkota KP, Bouras H, ElGharous M, Tychon B. Modeling genotype × environment × management interactions for a sustainable intensification under rainfed wheat cropping system in Morocco. Agriculture & Food Security. 2023;12(1):22.
24. Maryelle Kleyce MN, Fernandes GST, João V de NP, Rua ML, Santos MGM, Ribeiro LRT, et al. The application of machine learning to model the impacts of extreme climatic events on the productivity of dwarf green coconut trees in the Eastern Amazon. AgriEngineering. 2025;7(2):33.
25. Ming-Hwi Yao, Yung-Heng Hsu, Ting-Yi L, Chen Y, Chun-Tang L, Chi-Ling C, et al. Agricultural disaster prevention system: Insights from Taiwan's adaptation strategies. Atmosphere. 2024;15(5):526.
26. Moeletsi ME, Tsubo M. A move towards developing usable climate change adaptation and mitigation services for the agricultural sector. Climate. 2024;12(3):46.
27. Olowoyeye T, Abegunrin G, Sojka M. Are agroecosystem services under threat? Examining the influence of climate externalities on ecosystem stability. Atmosphere. 2024;15(12):1480.
28. Osumba JJL, Recha JW, Oroma GW. Transforming agricultural extension service delivery through innovative bottom-up climate-resilient agribusiness farmer field schools. Sustainability. 2021;13(7):3938.
29. Peng Y, Liu B, Zhou M. Sustainable livelihoods in rural areas under the shock of climate change: Evidence from China labor-force dynamic survey. Sustainability. 2022;14(12):7262.
30. Řehoř J, Trnka M, Brázdil R, Fischer M, Balek J, van der Schrier G, et al. Global hotspots in soil moisture-based drought trends. Environmental Research Letters. 2024;19(1):014021.