Abstract: Human agricultural activities are generating substantial carbon emissions across vast regions. Accurately understanding the current status, spatiotemporal patterns, and future trends of agricultural carbon emissions is crucial for optimizing carbon sequestration and emission reduction policies and enhancing climate adaptation capabilities. However, current methods for assessing agricultural carbon emissions face an inevitable challenge of reliance on statistical data, whereby regions with incomplete statistical records may encounter greater uncertainties in carbon accounting, analysis, and forecasting, or even lack results altogether. This limits our in-depth understanding of the spatiotemporal characteristics and future trends of agricultural carbon emissions in these regions. Taking Chongqing as a case study, this study integrated multi-source agricultural data combining government statistics and remote sensing monitoring within a county-level data framework to calculate agricultural carbon emissions from 2004 to 2023. Furthermore, spatiotemporal analysis methods including slope estimation, the Mann-Kendall test, Moran's I index, and the Getis-Ord Gi* index were employed to examine change characteristics, while prediction models based on ARIMA and three machine learning methods (Support Vector Machine, Random Forest, and XGBoost) were constructed to investigate trends. The results indicate that: 1) the multi-source data integration method for agricultural carbon emission accounting is feasible and reliable, yielding an average annual agricultural carbon emission of approximately 2.435 million tons for Chongqing. This result correlates significantly with accounting based on traditional statistical data (R²=0.932, p<0.05), compensating for missing county-level statistical data and demonstrating higher stability. 2) Significant source and regional differences exist in Chongqing's agricultural carbon emissions. Methane emissions from rice cultivation and carbon emissions from fertilizer use are the primary sources, with average annual emissions reaching 1.175 million tons and 0.809 million tons, respectively. The spatial agglomeration of agricultural carbon emissions in Chongqing has intensified year by year, with the global Moran's I index reaching 0.695, 0.615, and 0.64 in 2017, 2021, and 2023, respectively. Districts such as Wanzhou, Liangping, and Zhongxian were identified as emission hotspots, with average annual agricultural carbon emissions of approximately 0.106, 0.105 and 0.089 million tons, respectively; whereas Beibei, Yubei, and Jiangbei were identified as emission cold spots, with average annual emissions of approximately 0.03, 0.051 and 0.01 million tons, respectively. 3) The interpretable ARIMA-XGBoost prediction model performed well on an independent test set (R²=0.936) and revealed that Chongqing's agricultural carbon emissions will shift from a general stable state to a more widespread downward trend. Total emissions are projected to gradually decrease from 2.187 million tons to 1.788 million tons between 2024 and 2030. Factors such as rural employees, highway mileage and gross agricultural product exert a more significant influence on changes in agricultural carbon emissions, yet the spatially differentiated distribution across counties remains largely unchanged. This study highlights the advantages of information complementarity and reliability offered by the multi-source data integration method for agricultural carbon emission accounting, analysis, and forecasting, thereby enhancing the regional agricultural carbon emission assessment framework. It provides a scientific foundation for formulating carbon sequestration and emission reduction policies in Chongqing's agricultural and rural sectors and serves as a valuable methodological reference for selecting green and low-carbon development strategies in similar regions.