Global agricultural carbon emission efficiency: Using machine learning techniques to reveal driving factors and forecast future trends

Agricultural carbon emission efficiency (ACEE) is crucial for advancing global carbon neutrality goals. However, existing research at the national level often overlooks the function of agricultural carbon sinks and exhibits deficiencies in analyzing the driving mechanisms of ACEE and making precise predictions. To address this, this paper constructs a more comprehensive ACEE measurement system and introduces machine learning techniques to thoroughly analyze the spatio-temporal dynamics, driving factors, and future trends of global ACEE. Firstly, by incorporating agricultural carbon sinks as an ecological output, this study develops an ACEE measurement system covering 162 countries, overcoming the limitations of previous studies that were often confined to regional levels or neglected carbon sinks. Measurements based on the global super-efficiency Epsilon-Based Measure model reveal that from 1995 to 2021, ACEE generally increased across countries, but spatial differentiation intensified, exhibiting a significant Matthew effect. Secondly, this study combines interpretable machine learning and geographically and temporally weighted regression to unveil the driving mechanisms of ACEE from socio-economic, agricultural, and climatic dimensions. Agricultural production level is the primary driver for enhancing ACEE, and economic development level also demonstrates a significant promoting role. However, rainfall intensity and agrochemical use intensity are the main inhibiting factors. Urbanization level, industrial structure, and agricultural trade openness negatively affect ACEE in most countries, while the positive effects of technological progress have been diminishing annually. Finally, to enhance prediction accuracy, this study employs an optimized backpropagation neural network model to predict ACEE for different country groups from 2025 to 2035. The ACEE gap between high- and low-level country groups is projected to further widen, and the global divergence trend will become more pronounced.