Key points are not available for this paper at this time.
A novel, resource-efficient framework for a semi-online, pre-running database probabilistic flood forecasting system is presented in this manuscript. The system was designed for deployment in resource-constrained areas, with its application demonstrated through a case study in Semarang City, Indonesia. The substantial computational demands of traditional full-online numerical simulations, which are often prohibitive in developing countries, are circumvented by this approach. To achieve this, the framework utilizes pre-running databases built from historical rainfall, hydrologic, and hydraulic model data. It integrates daily calibrated probabilistic rainfall forecasts that are derived from a multi-model time-lagged ensemble analysis of outputs from the Global Forecast System (GFS) and Weather Research & Forecasting (WRF) models. This integration produces a daily probabilistic inundation map, valid for 24 hours with a 14-hour lead time, to assist decision-makers in assessing future uncertainty. The historical simulations of the model were found to exhibit good agreement with observational data, and a probabilistic rainfall forecast evaluation demonstrated a low Brier score, confirming its accuracy. While the model has acknowledged limitations, the framework represents a crucial step towards developing practical and accessible forecasting and prediction parts of flood early warning systems (FEWS) in similar regions.
Suwarman et al. (Thu,) studied this question.