A framework to quantify uncertainty of crop model parameters and its application in arid Northwest China

•We developed a framework that integrates sensitivity, uncertainty and parameter calibration.•Water stress parameters were more sensitive in severe drought than in full irrigation.•Times series measured data help MCMC iteration to get more reliable posterior distribution.•Interannual variation and severe water stress caused extra uncertainty of model residual error.•The framework improved the simulations under four irrigation scenarios in drought climate.