Wind downscaling is crucial for refining coarse‐scale wind estimates, improving local‐scale predictions, and supporting various applications like risk assessment and planning. Dynamic downscaling models demand extensive computational resources and time, leading to a shift toward more efficient statistical downscaling, whereas it often overlooks inter‐variable and inter‐station spatial correlations. Addressing this, we propose TerraWind, a deep learning‐based downscaling method for complex terrain regions. TerraWind enhances accuracy by incorporating topographic factors and inter‐station linkages, capturing wind field interactions with terrain at multiple scales. Experimental results in Eastern China demonstrate that TerraWind reduces wind speed Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) by an average of 42.6 and 33.3, respectively, compared to three interpolation methods (bicubic, bilinear, and Inverse Distance Weighting). Furthermore, TerraWind achieves an average reduction of 35.3 in wind speed MAE and 25.6 in wind speed RMSE compared to four deep learning models (Wind‐Topo, DeepCAMS, RCM‐emulator, and Uformer).

An analysis of tropical low formation in the Australian monsoon during January 2013 is presented based on European Centre for Medium Range Weather Forecast (ECMWF) analyses. The focus is on three lows that developed on the monsoon shear line, two of which eventually became tropical cyclones. One of the disturbances intensified as it made a loop over land. Interpretations of the formation are given in terms of vorticity dynamics. The roll up of low-level absolute vorticity associated with the shear line is a prominent feature of the early development of the lows, a mechanism akin to that believed to operate in the formation of hurricanes from Atlantic easterly waves. A notable feature of the two lows that eventually became tropical cyclones was the occurrence of persistent bouts of deep convection near the circulation centre. The development of the low that didn’t achieve tropical cyclone status was apparently thwarted by strong vertical shear and drying of the middle troposphere.

An analysis of numerical simulations of tropical low intensification over land is presented. The simulations are carried out using the MM5 mesoscale model with initial and boundary conditions provided by ECMWF analyses. Seven simulations are discussed: a control simulation, five sensitivity simulations in which the initial moisture availability is varied, and one simulation in which the coupling between moisture availability is suppressed. Changing the initial moisture availability adds a stochastic element to the development of deep convection. The results are interpreted in terms of the classical axisymmetric paradigm for tropical cyclone intensification with recent modifications.Spin‐up over land is favoured by the development of deep convection near the centre of the low circulation. As for tropical cyclones over sea, this convection leads to an overturning circulation that draws absolute angular momentum surfaces inwards in the lower troposphere leading to spin‐up of the tangential winds above the boundary layer. The intensification takes place within a moist monsoonal environment, which appears to be sufficient to support sporadic deep convection. A moisture budget for two mesoscale columns of air encompassing the storm shows that the horizontal import of moisture is roughly equal to the moisture lost by precipitation. Overall, surface moisture fluxes make a small quantitative contribution to the budget although, near the circulation centre, these fluxes appear to play an important role in generating local conditional instability. Suppressing the effect of rainfall on the moisture availability has little effect on the evolution of the low, presumably because, at any one time, deep convection is not sufficiently widespread.