Shiori OGURA

Error Characteristics of Short- Range Precipitation Forecasts and Their Application to Landslide Risk Assessment

Minjiao LU

In recent years, abnormal weather events have caused a significant increase in the frequency and severity of disasters induced by heavy rainfall, including sediment-related disasters. To implement effective disaster mitigation measures, it is essential to evaluate disaster risk in advance, particularly at an early stage before disaster occurrence. Therefore, the development of reliable early warning methods is strongly required.
Currently, the Soil Water Index (SWI) is widely used for sediment disaster risk assessment. However, the determination of warning thresholds requires substantial expertise and experience, which poses challenges for regional application and operational use. As an alternative, the Soil Moisture Deficit (SMD) has been proposed as a risk assessment indicator, as it allows simpler threshold setting and effective evaluation of sediment disaster risk. Moreover, because the temporal variation of SMD can be expressed solely by rainfall amount, the use of forecast precipitation enables risk assessment several hours in advance.
When short-term precipitation forecasts are applied, however, it is necessary to consider the uncertainty arising from forecast errors. Although previous studies have shown that one-hour-ahead forecast errors follow a normal distribution and that uncertainty-aware SMD can capture sediment disaster occurrence one hour in advance, error characteristics for longer lead times have not been sufficiently clarified.
This study aims to investigate the error characteristics of short-term precipitation forecasts for lead times of one to three hours and to evaluate their applicability to early sediment disaster risk assessment. Forecast error distributions were analyzed for each lead time, and SMD values considering uncertainty were calculated in an ensemble-based manner. Furthermore, sediment disaster risk was assessed using the probability of exceeding predefined warning thresholds.
The results demonstrate that incorporating forecast uncertainty into SMD-based assessment extends the applicable lead time and enhances the practicality of early warning methods for sediment disasters.