The impact of rainfall time series with different length in a landslide warning system, in the framework of changing precipitation trends

Background This study explores some possible impacts that climate change could have in regional scale landslide early warning systems based on rainfall thresholds. The early warning system of the Emilia Romagna region (Italy), was used as a case of study to assess how much the changing precipitation trends can affect the rainfall parameters used by the warning system and to assess whether the predictive power is higher when its thresholds are defined using long (60 years) rainfall time series or using a shorter (10 years) but more recent dataset. Results In our case of study, the use of recent but limited rainfall time series, although more representative of the contemporary rainfall regime, is a worse predictor compared to the longest possible rainfall datasets. The analysis of 60-years-long rainfall series facilitated the identification of extreme and anomalous events, as they better stand out from the average time series. Conclusions This study proves that early warning systems should not be considered as a static product: the rainfall thresholds should be regularly updated, the measuring network should be maintained, and the dismissal of oldest sensors should be carefully evaluated as their long time series improve the forecasting effectiveness or warning systems.