Taking Northwest China as the research object, the standardized precipitation evapotranspiration index(SPEI) and the standardized soil moisture index(SSMI) were used to characterize meteorological drought and agricultural drought respectively. Based on the three-dimensional recognition of drought events and the matching results of meteorological and agricultural drought events, the spatio-temporal and multi-angle response of agricultural drought to meteorological drought was established by using Copula function conditional probability distribution and Bayesian network probability model. The results show that based on the linear and nonlinear relationship models and the conditional probability distribution of Copula function, the optimal response relationship between agricultural drought and meteorological drought about drought duration, drought intensity, drought area and drought migration distance in Northwest China are quadratic polynomial model, nonlinear relationship model based on Frank Copula function, exponential model and linear model, respectively. The established response probability characteristic curve between meteorological and agricultural drought variables can quantitatively predict the probability of different degrees of meteorological drought developing into agricultural drought in Northwest China. Given the meteorological drought characteristics, the agricultural drought characteristics and its occurrence probability can be reliably predicted according to the optimal response relationship and response probability characteristic curve.