High spatiotemporal resolution typical meteorological year (TMY) data are essential for building energy modelling and urban climate studies. However, conventional TMY datasets, limited by sparse ground-based station coverage and infrequent updates, fail to meet the demands of detailed urban-scale simulations. To overcome these limitations, this study uses Australia as a case study and develops a new high-resolution dataset, the TMY derived from the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2), hereafter referred to as TMY-MER. A novel weather classification approach was introduced, utilizing a mean relative error index derived from the ratio of daily to monthly maximum solar radiation to identify clear-sky conditions. Uncertainty errors were spatially interpolated using the inverse distance weighting (IDW) method. The results reveal several limitations in the previously generated TMY datasets. TMY-MER demonstrates stable accuracy under clear-sky conditions, with annual average errors below 5 %, while under cloudy conditions, influenced by cloud simulation bias, errors can reach up to 50 %. Spatially, annual solar irradiance is overestimated by 30 % in southeastern coastal urban clusters, while errors in inland regions remain below 10 %. Temporally, the peak error during cloudy winter periods reaches 30 %, whereas summer clear-sky errors are under 5 %. Further analysis using the direct-diffuse separation model indicates a systematic overestimation of diffuse horizontal irradiance (DHI) within 6 %, and an approximately 20 % negative bias in direct normal irradiance (DNI). Validation through building cluster simulations shows that the optimized dataset achieves over 90 % consistency with traditional TMY data, with monthly mean errors below 5 %. The multidimensional error assessment framework significantly enhances the reliability of reanalysis data for use in complex climate zones, supporting dynamic energy system planning and urban thermal environment modelling.