Toru KOBAYASHI

Error Characteristics of  Very-Short-Range Forecasting and Its Application to Landslide Risk Assessment

Minjiao LU

In 2023, Lu et al. presented that soil moisture deficit is a good indicator for landslide risk assessment.
This announcement suggests that the use of predicted rainfall several hours ahead in calculating soil moisture deficit in the event of a disaster may enable landslide risk assessment several hours into the future, leading to early warning of a landslide disaster.
In this study, we investigate the possibility of assessing the risk of landslide disaster one hour ahead by estimating the SMD one hour ahead using the forecasted rainfall from the Very-Short-Range Forecasting as the forecasted rainfall. However, there is a problem that has not yet been clarified in general, which is the need to consider the error characteristics of the forecast results of the Very-Short-Range Forecasting when using the forecast rainfall.
In this study, we first conducted a simple survey of the error characteristics of forecasts based on Very-Short-Range Forecasting , and then made an ensemble forecast of SMD one hour ahead, taking into account the error characteristics. The probability of exceeding an arbitrary criterion for issuing a disaster warning is calculated from the predicted results, and the possibility of early warning of disasters by estimating the SMD one hour ahead is investigated.
As a result, for the data obtained in this study, the error characteristics of the forecast results from the Very-Short-Range Forecasting tended to follow a normal distribution. The reliability of the risk assessment of landslide disaster one hour ahead under the assumption of a normal distribution is higher than the arbitrary disaster capture rate, indicating that early warning of disaster is feasible in the current situation.