1 Department of Building, Civil, and Environmental Engineering, Concordia University, 1455 de Maisonneuve Blvd. W. Montreal, Quebec, Canada H3G 1M8. Electronic address: darafzal@outlook.com.
2 Department of Building, Civil, and Environmental Engineering, Concordia University, 1455 de Maisonneuve Blvd. W. Montreal, Quebec, Canada H3G 1M8. Electronic address: zhichen@bcee.concordia.ca.
3 Faculty of Agriculture, University of Granma, Granma, Cuba. Electronic address: sfrodriguez1964@gmail.com.
4 Department of Building, Civil, and Environmental Engineering, Concordia University, 1455 de Maisonneuve Blvd. W. Montreal, Quebec, Canada H3G 1M8. Electronic address: fariborz.haghighat@bcee.concordia.ca.
5 Meteorological Center of Granma Province, Granma, Cuba. Electronic address: bettyrosales95@gmail.com.

Assessing the impact of greenhouse gas (GHG) emissions on agricultural soils is crucial for ensuring food production sustainability in the global effort to combat climate change. The present study delves to comprehensively assess GHG emissions in Cuba's agricultural soil and analyze its implications for rice production and climate change because of its rich agriculture cultivation tradition and diverse agro-ecological zones from the period of 1990-2022. In this research, based on Autoregressive Distributed Lag (ARDL) approach the empirical findings depicts that in short run, a positive and significant impact of 1.60 percent % in Cuba's rice production. The higher amount of atmospheric carbon dioxide (CO₂) levels improves photosynthesis, and stimulates the growth of rice plants, resulting in greater grain yields. On the other hand, rice production index raising GHG emissions from agriculture by 0.35 % in the short run. Furthermore, a significant and positive impact on rice production is found in relation to the farm machinery i.e., 3.1 %. Conversely, an adverse and significant impact of land quality was observed on rice production i.e., -5.5 %. The reliability of models was confirmed by CUSUM and CUSUM square plot. Diagnostic tests ensure the absence of serial correlation and heteroscedasticity in the models. Additionally, the forecasting results are obtained from the three machine learning models i.e. feed forward neural network (FFNN), support vector machines (SVM) and adaptive boosting technique (Adaboost). Through the % MAPE criterion, it is evident that FFNN has achieved high precision (91 %). Based on the empirical findings, the study proposed the adoption of sustainable agricultural practices and incentives should be given to the farmers so that future generations inherit a world that is sustainable, and healthy.