This paper constructed a dataset of China’s abandoned farmland from 2000 to 2020, and analyzed the evolution characteristics and leading factors of the spatial pattern. The findings are as follows. 1) From 2000 to 2020, the abandoned area and abandonment rate showed a continuous growth trend, with the abandoned area increasing from 255.96×10⁴ hm² to 3113.03×10⁴ hm², and the abandonment rate increasing from 1.29% to 14.08%. The areas with a higher abandonment rate were concentrated in the west side of the Hu Line. 2) There was a significant spatial autocorrelation of farmland abandonment. The “high-high” agglomeration area continued to expand and was concentrated on both sides of the Hu Line, while the “low-low” agglomeration area was mainly distributed in the eastern plain area. Meanwhile, as the abandonment rate continued to rise, a phenomenon of polarization emerged within agricultural regions. 3) In 2020, five factors, including soil type, population density, elevation, rural population proportion and geomorphic type, had strong explanatory power to the spatial differentiation of farmland abandonment. The results of two-factor detection showed nonlinear or double-factor enhancement, and the synergistic driving effect within the natural environmental factor group was stronger. On the basis of natural environmental factors, the explanatory power of socio-economic factors was obviously enhanced.