A tropical cyclone rainfall climatology-persistence model for the Taiwan area

Due to the central mountain range with altitude up to about 3 km, the amount and distribution of rainfall in Taiwan associated with typhoons or tropical cyclones (TCs) depends both on the distribution of convections within the TCs (internal structure) and the orographic effect. Since 1989, the spatial and temporal characteristics of TC rainfall that affected Taiwan have been well observed by the 371 local rain gauge stations available as well as radar reflectivity data from the local Doppler radar network set up in recent years. The fact that there exist different characteristic rainfall patterns for TCs that approached the Taiwan area from different directions allows application of statistical methods to forecast TC rainfall before the skill of dynamical models becomes stable and good enough. The development of a TC rainfall climatology-persistence (R-CLIPER) model for the Taiwan area is described. A preliminary version of the model that utilizes rainfall climatology and 6-h persistence only has been used in Taiwan's National Center for Disaster Reduction in the 2004 and 2005 TC seasons for rainfall estimation up to a lead time of three days (Lee et al. 2005). Except a general trend of underestimation of the maximum rainfall areas that is common in climatological models, the correlation between the R-CLIPER forecasts and observations is about 0.7|0.8 (0.5|0.6) in the 3-6-h (24-h) forecast range in all river basins based on independent verifications. However, the overall equitable threat score received by the model for forecasts up to 24 h is only around 0.2 for smaller threshold and lower for higher threshold. Moreover, the skill is not very sensitive to the height of the station. This indicates that this preliminary R-CLIPER model can capture the overall rainfall pattern quite well especially in the short forecast range but its ability to predict the torrential rain is not adequate. This study discusses how the skill of the R-CLIPER model is improved by incorporating TC-structure related predictors and environmental parameters that may affect the rainfall pattern. The model will also serve as a skill reference for forecasts from dynamical models.

Reference: Lee, C.-S., L.-R. Huang, H.-S. Shen, and S.-T. Wang, 2005: A climatology model for forecasting typhoon rainfall in Taiwan. Natural Hazards (in press)