Search

The accurate forecasting of agricultural carbon emissions is essential for formulating strategies to achieve carbon peak and neutrality objectives within the agricultural sector. However, existing methodologies for predicting agricultural carbon emissions have notable limitations. To address these shortcomings, Shanghai farm was considered as a case study to conduct research utilizing a neural network approach. Agricultural carbon emissions from the Shanghai farm from 2011 to 2021 were computed using the emission-factor method. Subsequently, a Back Propagation （BP） neural network model was developed to predict carbon emissions, employing the GDP of the planting, animal husbandry, and fishery sectors as input variables. The model was further improved through the application of an optimized sparrow search algorithm, which was then employed to forecast the future carbon emissions of the farm. The results show that the BP neural network improved via the optimized sparrow search algorithm demonstrated a prediction accuracy of 96.14%, a root mean square error （RMSE） of 12 100 t·a -1 and a correlation coefficient （ R 2 ） of 0.995 2. These metrics underscored the superior performance of the enhanced model. Compared with the multiple running results of pre-improved models, the neural network improved by the optimized sparrow search algorithm enhanced both the accuracy and stability of carbon emission prediction significantly, with the prediction accuracy consistently approaching approximately 95%, the root mean square error remaining below 20 000 t·a -1 , and the correlation coefficient exceeding 0.99. Predictive analysis of future carbon emissions from the Shanghai farm indicated a predominant contribution from the animal husbandry sector to the total carbon emissions, suggesting that effective management of the scale of animal husbandry operations could significantly mitigate overall carbon emissions.