Droughts pose significant threats to both natural ecosystems and human society. Understanding how meteorological droughts evolve into agricultural droughts is critical for developing effective mitigation strategies. We used Enhanced Convergent Cross Mapping (ECCM) alongside Artificial Neural Network (ANN), Deep Neural Network (DNN), and Feed Forward Neural Network (FFNN) models with Explainable Artificial Intelligence (XAI) to identify and analyse nonlinear drought propagation pathways of China-Pakistan Economic Corridor (CPEC) over 1981-2022. We quantified causal relationships and key influencing factors, focusing specifically on drought propagation counts (DPCs) and the drought propagation intensity index (DIP). Our findings showed that: (1) Approximately 73 % of monitoring stations exhibited higher maximum Pearson Correlation Coefficient (PCC) than ECCM values, with the strongest correlations and causalities observed in the western regions (PCC: 0.48-0.79; ECCM: 0.36-0.77), and the weakest in eastern areas (PCC: 0.06-0.33; ECCM: 0-0.25). Higher quantiles (8-10) showed stronger agreement between maximum PCC (MPCC) and maximum ECCM (MECCM), while lower quantiles (<3) indicated more complex drought propagation dynamics with a strong linear fit between MPCC and MECCM (R-2 = 0.87); (2) Based on ECCM, propagation times ranged from 3 to 4 months in northern/southern Xinjiang and eastern Pakistan to 1-2 months in eastern Xinjiang and western Pakistan, whereas PCC-based methods extended this range to 5 months in some regions. Southwestern Pakistan and northwestern Xinjiang exhibited the highest propagation rates (77.95 %), whereas eastern regions showed weaker transitions (5.64 %); (3) XAI-based interpretations of ANN, DNN, and FFNN models identified maximum temperature and soil moisture as the most significant predictors, achieving high predictive accuracies for DPCs (R-2 > 0.74) and DIP (R-2 > 0.68). Interactions among the climate variables explained regional nonlinearities in drought propagation. These findings offered valuable insights for drought management in arid and semiarid regions.