Predicting methane and nitrous oxide emissions from Indian cattle farming using advanced time series techniques

This study aims to forecast methane (CH₄) and nitrous oxide (N₂O) emissions from cattle rearing in India, which contribute significantly to agricultural greenhouse gas (GHG) emissions. Data on these emissions was collected from the Food and Agricultural Organization for the years 1961–2022. Three time series models, namely, exponential smoothing (Holt-Winters), autoregressive integrated moving average (ARIMA), and trigonometric seasonality, Box-Cox transformation, ARMA errors, trend, and seasonal components (TBATS) were employed to predict future emissions. The dataset was partitioned into training (1961–2012) and testing (2013–2022) sets to evaluate model performance. Diagnostic metrics, including Akaike Information Criterion, root mean square error, mean absolute percentage error, and mean absolute scaled error, were used to assess accuracy. Results indicated that the ARIMA model outperformed the other 2 forecasting models by making over 90% accurate predictions. For N₂O, ARIMA (0,1,0) was identified as the optimal model, while ARIMA (2,1,2) was selected for CH₄. Thus, the study validates the use of ARIMA model in GHG forecasting. The study projects emissions up to 2030, providing critical insights for policymakers to design targeted mitigation strategies. The study also presses the need for implementing sustainable cattle management practices for cutting emissions in India.