Workflow to generate TPRR

The method that subdivides the long-term continuous integration into short ones can alleviate the growth of systematic errors in long-term integrations, which is the so-called re-initialization. Maussion et al.³⁸ and Wang et al.⁴² conducted HAR and HARv2 using re-initialization, and both datasets show good performance in describing the climate over the TP. In this study, the re-initialization strategy was adopted, and the spectral nudging (SN) method⁷⁴ which is considered as an indirect assimilation method and can prevent the simulation from drifting away from the forcing, was applied to wind fields above planetary boundary layer (PBL). The workflow for producing TPRR is presented in Fig. 2. Each short run starts from 12:00 UTC and is integrated for 36 hours, with the first 12 hours treated as spin-up time and the remaining 24-hour output combined into the long-term TPRR. For the potential users of TPRR (atmospheric scientists, hydrological scientists, etc.), post processing is applied to the output of WRF for ease of use. TPRR provides common surface and upper-air isobaric meteorological variables, separated by per variable, per year and per time aggregation (hourly and daily).

Workflow of TPRR product, including each short-term run (b) and their combination (a).