This study conducts a long-term climatological analysis and future projection of winter precipitation patterns in Yilan, Taiwan. By analyzing 40 years (1981-2020) of gridded daily precipitation data, we noticed that the rainfall hotspot in Yilan was located in the southern mountainous region and decreased toward the Lan-Yang Plain. We used a cloud-resolving numerical model - TaiwanVVM simulation outputs to understand the interaction of upstream background wind and terrain effect, which induced local circulation variations and affected rainfall distributions. We note that when the easterly component of the background wind was stronger, the convection became more significant in the southern mountainous areas of Yilan.
We analyzed gridded precipitation data using a hierarchical clustering method. We grouped Yilan's rainfall distribution into five clusters. We found that when the upstream wind field is dominated by northerly winds with a low mixing ratio, precipitation in Yilan decreases. The rainfall was concentrated in the southern mountainous areas and the northern coast. Conversely, heavy rainfall events are more likely when the upstream wind field shifts to easterly winds with high moisture content, with precipitation hotspots expanding from the southern mountainous areas to the Lanyang Plain. Therefore, we noticed that precipitation intensity and frequency variations are strongly related to moisture transport in the upstream lower-level wind fields.
We used the K-Nearest Neighbor algorithm to project future rainfall features and combined one of the CMIP6 global climate models, TaiESM1, with simulation outputs. Despite significant global warming scenarios, we observed that heavy rainfall events in Yilan have not generally increased. Instead, the changes in rainfall patterns vary under different warming scenarios and time periods. Under the warming scenario (SSP1-2.6), the occurrence frequency of the significant rainfall clusters will be decreased across all periods. In the warming scenario (SSP2-4.5), precipitation frequency increases in the "near future period" and "end of the century" but decreases significantly in the "middle of the century." In the scenario (SSP5-8.5), frequencies increases in the " near future period ".
