Rainfall thresholds for landslide occurrence: systematic underestimation using coarse temporal resolution data

Abstract

Rainfall thresholds for landslides occurrence derived in real applications tend to be lower than the ones one would obtain using exact data. This letter shows how the use of coarse temporal resolution rainfall data causes a systematic overestimation of the duration of the triggering rainfall events that directly contributes to thresholds underestimation. A numerical experiment is devised to quantify this systematic effect for the relevant case of power-law depth/intensity–duration thresholds. In the examined conditions, i.e., the frequentist method at 5% non-exceedance probability level, ~ 70% underestimation of the scale parameter and ~ 60% overestimation of the shape parameter of the thresholds is to be expected using daily resolution rainfall data, but the exact quantification depends on the specific characteristics of each study case. The underestimation increases as the temporal resolution becomes larger than the expected minimal duration of the triggering events. Under operational conditions, sensitivity analyses based on the methods and datasets of interest are advised.