Analog forecasting of tropical cyclone rainfall in the Philippines

Abstract

Tropical cyclone (TC) rainfall results in damages to properties and loss of lives. It is also a significant source of freshwater in the Philippines. This study describes a method in forecasting accumulated TC rainfall using analogous TCs from historical datasets. A TC rainfall database where precipitation within 5∘ of TC centers was created for all landfalling TCs from 1951 to 2015. To predict TC rainfall, the mean rainfall of all past TCs with similar tracks included in the database, referred to as analog TCs, is calculated. Landfalling TCs from 2016 to 2018 are used to optimize the selection of past analog TCs. Each past TC member was also adjusted according to a target TC's intensity and movement speed. The optimized analog method is then applied to landfalling TCs from 2019 to November 2020. Results show that the composite rainfall from past TCs within 1.8∘ of the forecast TC yields the best hit rate of intense rainfall. The analog TC rainfall forecast generally has a similar spatial distribution as the observed TC rain. However, this method tends to miss extreme rainfall values due to a “smoothing” effect caused by the variability of extreme rain locations of each TC member and constraints in the rainfall data used in the database. Nevertheless, forecast assessment results show that analog TC rainfall forecasting performed better than the WRF model in predicting intense and inland rainfall. In addition to it being computationally inexpensive, it can complement the inherent biases of dynamical models.