摘要:Abstract This study investigates the impact of uncertainty in initial conditions on 24-h sea-level pressure predictions near 0509 Typhoon Matsa by using conditional nonlinear optimal perturbation (CNOP). The CNOP is calculated by using a newly proposed fast algorithm. The model used is the Global/Regional Assimilation and PrE-diction System (GRAPES). The sensitivity of the 24-h predictions is studied in terms of horizontal and vertical ranges and also in terms of different initial state variables. To study the sensitivity of 24-h sea-level pressure predictions to different initial state variables, four functions are given as metrics to find the sensitive initial locations. The results show that the main prediction errors come from initial uncertainty on the levels below 200 hPa and at the region south of about 37.6°N, with more sensitivity to initial winds than to other initial state variables.