A Hybrid Mathematical Models for Predicting Global Climate Change

The industrial revolution marked the beginning of modernization in human civilization, and also marked the sharp rise in greenhouse gas emissions and global temperatures. To better understand trends in global climate change, we aim to utilize data on carbon dioxide levels and land-ocean temperatures to learn past trends and predict future changes. First, the CO2 concentration dataset, using statistical methods, is analyzed and visualized. From the statistical summary and graphs, it can be concluded that the global CO2 level has been constantly increasing since the 1960s. Based on the dataset, three models were constructed to analyze the changing trend of CO2 levels in the past and extrapolate the future: Autoregressive Integrated Moving Average (ARIMA), grey forecast, and a more refined prediction model that considers factors affecting CO2 levels with Long Short Term Memory (LSTM). All three models disagree that the CO2 level will reach 685 PPM by 2050. And each model predicts CO2 level of 685 PPM will be reached by the end of the century and when. Afterward, the pros and cons of the models are compared. Second, the model of the changes in global land-ocean temperature is constructed. ARIMA is used to model and predict the upcoming temperature and the time when it is going to reach certain designated points. Pearson's correlation shows a strong correlation between global temperature and CO2 level. Hence, these two variables are modeled with linear regression. However, the regression-based predictions did not match the forecast from earlier models, so an refined model incorporating more variables and perspectives was built. The refined model is a more bottom-up approach. It quantifies the radiative forcing of individual factors and makes predictions based on the predicted outcomes of each factor. The model predicts the temperature difference of 3.55 °C from the base period, 1.25 °C in 2031, 1.5 °C in 2039, and 2 °C in 2052.