Effects of the assimilation of relative humidity (RH) reproduced from dropsonde data from the Tropical cyclones‐Pacific Asian Research Campaign for Improvement of Intensity estimations/forecasts (T‐PARCII) campaign and Himawari‐8 satellite data on the simulation of Typhoon Lan (2017) were investigated herein using a weather research and forecasting model. The lateral boundary and initial conditions were obtained from Global Forecast System (GFS) forecast data. Four experiments varying from the initial vortex, assimilation of the reproduced RH (ARH), and ocean model (OCEAN, three‐dimensional Price‐Weller‐Pinkel upper‐ocean circulation model) activation were conducted for 42 h to evaluate tropical cyclone (TC) forecast performance: the GFS experiment and dynamical initialization (DI) experiments such as DI, DI‐OCEAN, and DI‐ARH‐OCEAN, respectively. All track forecast errors were less than 100 km until landfall, that is, up to 36 h. TC intensity forecasts such as the minimum sea‐level pressure and maximum surface wind speed were slightly improved in DI‐related experiments compared to the GFS experiment. The DI‐ARH‐OCEAN experiment, in particular, demonstrated improvements in both TC forecasts and convective areas. The ocean‐coupled experiments yielded significant sea surface temperature cooling in the rear‐right quadrant of TC, forming a stable boundary layer that could suppress the convective activity, particularly in the lower troposphere. These findings support that compared to the original DI method, ARH could improve initial conditions, resulting in more accurate TC forecasts. Furthermore, it may argue the necessity and urgency of regular aircraft surveillance of TCs in the western North Pacific area.