The chapter examines the increasing global concentration of carbon dioxide
(CO2
) and its implications for climate change. It analyses CO2
level shifts over the ten
years spanning from 2014 to 2024. The dataset obtained from Kaggle synchronizes
year, month, day, cycle, and trends. This paper aims to determine the CO2
cycle and
trend for the next year using five different forecasting models: Seasonal Autoregressive
Integrated Moving Average (SARIMA), Prophet, Long Short-Term Memory networks
(LSTM), Extreme Gradient Boosting (XGBoost), and Exponential Smoothing state
space model (EST). Model evaluation has been achieved by measuring the following
indicators: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Rsquared (R2
) values. In contrast, the LSTM model produced accurate forecasts with
minimal errors, attaining an MAE of 0.09, RMSE of 0.12, and a high R2
value of 1.00.
SARIMA’s forecasting ability, on the other hand, was reported to be weak (MAE:
11.06, R2
: -16.10), whilst Prophet and XGBoost models produced average MAEs of
0.43 and 1.66, respectively, and the R2
values of 0.97 and 0.30, respectively. The ETS
model recorded an MAE of 5.64 and an R2
of -3.69. These measuring indicators
highlight the effectiveness of the LSTM model, which allows the prediction of trends
and cycles in CO2
emissions and can be particularly useful for climate policies and
researchers. This chapter also highlights the need for advanced modeling to inform
effective policies against the impacts of climate change.
Keywords: CO2 Cycle, ETS, Forecasting, LSTM, MAE, Prophet, RMSE, R-Square, SARIMA, Trend, XGBoost.
