Evaluation of rainfall prediction in the vicinity of Solomon Islands with a high-resolution non-hydrostatic model

Abstract

Meteorological authorities in many developing countries can benefit from running a regional model optimized for their countries in addition to rainfall predictions provided by major centers in developed countries. As one such activity, Solomon Islands Meteorological Services (SIMS) recently started to operationally run their regional model based on the non-hydrostatic model of Japan Meteorological Agency (JMA). This study evaluates the skill of rainfall predictions by the SIMS operational model with a 15-km mesh (SI15) based on 219 cases of 36-hour forecasts initialized every five days during 2017-2019. We also conducted runs with a 1.875-km mesh (SI01) in anticipation of future numerical resources. The skill was compared with the forecasts of the JMA Global Spectral Model (GSM), using rain gauge data and satellite-based data. The forecasts by SI15 and SI01 exhibited better performance than the GSM in terms of bias and threat score, with SI01 performing the best. Additionally, SI01 successfully reproduced the diurnal variation where precipitation becomes strong over land in the evening and on the eastern side of the Solomon Islands over the sea in the early morning. In contrast, SI15 showed weak biases around the initial time and strong biases after 18 hours of forecast time.