Dynamic Filtering in Lagrangian Model‐Based Moisture Source–Receptor Diagnostics
Moisture source–receptor diagnostics based on Lagrangian moisture tracking models are essential numerical tools for understanding global and regional hydrologic processes, valued for their computational efficiency and spatial accuracy. Among these diagnostic methods, WaterSip is the most widely used one, with its reliability further enhanced by the bias correction capabilities of the Heat And MoiSture Tracking framework (HAMSTER). However, these methods predominantly employ uniform humidity and altitude thresholds for identifying precipitation and evaporation moisture, limiting their adaptability to diverse surface properties and climatic conditions globally. This study proposes a dynamic filtering (DF) method that employs spatiotemporally varying relative humidity (RH) and altitude thresholds to improve the trajectory selection and moisture identification. The DF method enables the specific humidity changes of selected particles to more faithfully represent precipitation loss and evaporation gain during moisture transport, and enhances the recognition of spatial contrasts in moisture contributions over source regions with complex land–sea distributions. Comparison with the Weather Research and Forecasting model with Water Vapor Tracers (WRF‐WVTs) indicates that the DF method brings WaterSip simulations into closer agreement with the WRF‐WVTs results, showing potential to reduce the systematic overestimation of nearby source contributions and underestimation of distant contributions. In a case study over the Tibetan Plateau (TP), the DF method highlights the contribution of westerly moisture sources to precipitation in the region. Nevertheless, the standalone DF method may still retain biases in total tracked moisture, necessitating subsequent bias correction (e.g., HAMSTER) to adjust overall moisture amounts.