Abstract: Using the data set of 189 meteorological stations in the Yangtze River Basin from 1963 to 2018, the stepwise multiple regression (SMR) models are established for monthly (seasonal) average air temperature, monthly (seasonal) average maximum and minimum air temperatures to analyze the variation characteristics of near-surface air temperature gradients (TG). The forcing processes for TG variations are then investigated by using precipitation gradient (PG), relative humidity and moisture flux. The results are as follows: (1) The latitudinal (TGa) and longitudinal (TGo) gradients for respective classes of temperatures are steeper in winter but shallower in summer, while the changes of vertical gradients (TGe) are different for three classes of temperatures. Moreover, there are steeper values of TGa and TGe in the nighttime than daytime. (2) Although the partial correlation coefficients between temperature versus latitude, longitude and elevation are greater than the corresponding TG values, their changing trends are similar. All classes of temperatures have the highest partial dependence on elevation, while the average minimum temperature has overall the strongest dependence on each of geographical factors. (3) The space-time variations of relative humidity in the basin make the SMR models of air temperature have a more powerful performance in summer than winter. However, the similar geographic environments between the stations within the same altitude range lead to the model residuals exhibiting overall positive or negative values in a season. (4) The TG changes are modulated and controlled by the topography, precipitation, relative humidity and moisture flux in the basin. The seasonality of TGa variations is intensified by the conversion of moisture flux, and the stronger the dry cold air is, the steeper of the TGabecome. Meanwhile, the shallower values of TGomay be ascribed to the relatively uniform distribution of relative humidity, while the terrain effect makes them steeper. Besides, there is an opposite trend between vertical gradients of average maximum and minimum temperatures owing to the different temperature effects of cloud-radiative forcing in the day and night time.