Evaluation and Implementation of Various Bayesian approaches to model predictions of future climate change 1 Saswata Lahiri Department of Data Science and Business Systems SRM Institute of Science and Technology Katthankulathur, Chennai, India sl9039@srmist.edu.in 2 Aagam Shah Department of Data Science and Business Systems SRM Institute of Science and Technology Katthankulathur, Chennai, India as4094@srmist.edu.in 3 T Veeramakali Department of Data Science and Business Systems SRM Institute of Science and Technology Katthankulathur, Chennai, India veeramat@srmist.edu.in Abstract— Climate change is a major environmental issue that affects the entire planet. Accurate and reliable predictions of future climate change are essential to inform policy decisions and to develop effective mitigation and adaptation strategies. Bayesian modeling techniques have been shown to be effective in predicting future climate change by combining different sources of information, including historical data, climate models, and expert knowledge. This project aims to evaluate and implement various Bayesian approaches to model predictions of future climate change. The project will involve web scraping for data acquisition and preparation, model development using Bayesian methods, model evaluation, and implementation of the models. The project will use different sources of data, including historical climate data, climate models, and expert knowledge. The results of the project will help to improve the accuracy and reliability of climate change predictions and will inform policy decisions and the development of mitigation and adaptation strategies. The use of web scraping techniques will allow for the acquisition of a large and diverse set of data, enhancing the robustness and accuracy of the models. Keywords—Bayesian Neural Network, Machine learning, Bayesian Approaches, climate change prediction I. INTRODUCTION Research by various organizations have mostly stated that climate change will always have an adverse effect on the environment and our lives as long as we do not take proper measures to protect our environment. Human activities are responsible for increasing effects of global climate change round the world. The effect of world wide climatic changes have been felt in every a part of the globe. consistent with United Nations Framework Convention on global climate change (UNFCCC), Asia, Africa and Latin America are among the regions of the planet, which have severely been suffering from the scourge. Using BNNs in global climate change prediction involves training the network on historical climate data then using it to predict future climate conditions. The advantage of BNNs is that they will incorporate prior knowledge and expert opinions, which may improve the accuracy of the predictions. Additionally, the uncertainty estimation provided by BNNs allows decision-makers to form informed decisions supported the extent of confidence within the predictions. II. LITERATURE SURVEY [1]In this paper, they evaluated the performance of 3 climatic models by comparing their correlation pattern and Root Mean Squared Value. In order to replicate rainfall and temperature in Asia Pacific, CSIRO, HadCM3, and CCMA were prepared. Hindcasting was employed to put the mathematical model to the test. To test if the output of the model accurately reflects the known outcomes, all estimated inputs of any historical occurrences have been added into the model. This conceptual model was used to assess how vulnerable Canadian water fish habitats are to the effects of climate change. To evaluate the anticipated effects of global climate change on aquatic habitats in the Asia Pacific, a similar framework is frequently created. [2] They presented posterior inferences using four subsets of the Jones data which were corresponding to the different time periods mentioned in the paper. The clearly successively shorter periods of time have been suggested in the study since any trend or pattern in the global climate change are much more prominent and clear in the latter time period. [3] The 5–95% percentiles of the probabilistic forecasts of the worldwide mean SATs are represented by the annual time series of multi-model weighted ensemble means. BF and EM's mean values are quite comparable to one another. They are greater than AEM, with a maximum difference of roughly 0.3 K. The PDFs in the upper tail are owing to the Bayesian weighting, as seen by the three forecasts' 5% percentiles being extremely near to one another and the 95% percentiles of BF and EM being bigger than those of AEM. [4] the mixture of analyses they presented here provides sufficient proof to question the notion of cooling climate being the first explanation for the population decline observed during the Late and Final Chulmun periods. If the cooling of the climate has been directly liable for effecting or causing the Chulmun population to