Evaluation of agricultural carbon emissions in Xinjiang and analysis of driving factors based on machine learning algorithms

Agriculturl carbon emissions are the second-largest source of carbon in the world. Therefore, clarifying the patterns of agricultural carbon emissions is crucial for achieving carbon peaks and neutrality. To explore the law of agricultural carbon emissions in Xinjiang and promote agricultural carbon emission reduction, agricultural carbon emissions in Xinjiang were measured based on carbon emission coefficients published according to the carbon emission links generated in the process of agricultural production. Furthermore, spatial correlation models, such as the Moran and learned index structure for spatial data (LISA) indices, were used to measure the spatial clustering patterns of agricultural carbon emissions in Xinjiang. A random forest machine learning model was then used to quantitatively analyze the factors influencing the efficiency of agricultural carbon emissions. The results indicated that: 1) agricultural carbon emissions grew slowly from 2010 to 2019, from 292.24×0⁴ t to 379.69×10⁴ t, with an average annual growth rate of 3.33%. 2) Applications of chemical fertilizers and agricultural films were the main sources of agricultural carbon emissions in Xinjiang, accounting for 58.06% and 39.03%, respectively. 3) Xinjiang’s agricultural carbon emission efficiency increased steadily, with a faster growth from 2010 to 2013 and a slower growth from 2014 to 2019. The main distribution range of carbon emissions efficiency increased from less than 50 ¥∙t⁻¹ to 50–100 ¥∙t⁻¹. 4) The agricultural output values in the high-high agglomeration areas of Xinjiang with high agricultural carbon emission efficiency were relatively low because of the low material input. In contrast, the agricultural output values in the low-low agglomeration areas were relatively high, however, where the level of technology and management was low, and the material input was extremely high. The efficiency of agricultural carbon emissions in Xinjiang has room for improvement. 5) Overall agricultural carbon emission efficiency was higher in the southern region with lower precipitation, whereas the northern region with higher precipitation exhibited moderate emissions. Precipitation may indirectly affect agricultural carbon emission efficiency by affecting the level of agricultural development and production technology. 6) Carbon emission efficiency decreased sharply with increased agricultural scale when the agricultural scale was between 0.12 and 2.02 hm² per person. Moreover, the influence on agricultural carbon emissions efficiency decreased when the agricultural scale exceeded 2.02 hm² per person. There was a significant negative effect on agricultural carbon emission efficiency when cultivated land was between 120 and 17 220 hm². In contrast, its’ effect on agricultural carbon emission efficiency was more moderate when cultivated land was larger than 17 220 hm². Rural economic development level had a positive effect on carbon emission efficiency. Furthermore, carbon emission efficiency exhibited a “U” shaped pattern as a function of agricultural electrification degree. Comprehensively considering the two aspects of improving agricultural output value and agricultural carbon emission efficiency, the degree of agricultural scale and the scale of arable land should be further improved to increase agricultural output value, and the level of rural economic development and the degree of agricultural electrification should be further improved to increase the efficiency of agricultural carbon emissions.