Using in situ observations and numerical modeling, this study investigates the dynamical mechanisms of seasonal variability of water temperature in the Yellow Sea (YS). Observations indicate that bottom temperature lags 3--4 months behind surface temperature in reaching a maximum in the central YS. Wave-tide-circulation coupled model simulates this time lag and indicates that the diffusion process is a key factor governing the temperature variation below the surface layer. Based on the diffusion equation of temperature, a scheme is developed to estimate the vertical diffusion coefficient. At an observation station located at 36¿00'N 124¿00'E, the diffusion coefficients from April to September are estimated by using the temperature data from 1954 to 1985. The mean diffusion coefficient (MDC) in the upper layer from 0 m to 15 m is almost one order of magnitude larger than those in the middle layer from 20 to 40 m, except in April. In the middle layer, the MDC is inversely proportional to the squared buoyancy frequency, and the mean value of MDC averaged from June to September is 0.28 cm2 s-1. The inverse proportionality agrees with the Osborn's relation, which has been used to estimate the diapycnal diffusivity.