Abstract
The denudation of young reliefs prone to landslides can have severe consequences for society and the environment. However, landslide databases and the additional information (landslide type, date and triggering factors) necessary to deal with landslide hazard assessment and the development of effective and reliable landslide warning systems are usually scarce or non-existent. In this way, by taking into account the date of landslide events and by expanding the analysis of cumulative rainfall from these dates to a broader time period that includes the days or months leading up to a landslide, the corresponding triggering rainfall threshold can be assessed more accurately. In this paper, a methodology based on a partial duration series analysis applied to rainfall variables allows the possibility to better understand precipitation patterns. Another advantage of analysing precipitation variables within a broader time period is the ability to identify greater accuracy rainfall anomalies such as extreme rainfalls with their return period related to a low number of dated landslide events (in this case, 20 landslide events). The landslide spatial distribution within a regional area requires the processing and analysis of data from multiple long-term historical daily rainfall records from different rainfall gauges, which notably increase the number of calculations to be dealt with. To overcome this inconvenience, these processes were streamlined by using macro-automation. Additionally, different rainfall durations can be interactively identified from graphical outputs that show anomalies on more than one rainfall variable after applying this methodology. Among these rainfall variables, the antecedent accumulated rainfall (A1) was found to be the most suitable to apply the occurrence probability analysis. When compared to other variables, the return period values of A1 were determined to be conservative, neither too high nor too low. Using this approach, the return period curve was shown to be an important graphic object in detecting uncommon rainfalls that are contemporaneous or previous to landslides. The relevant findings of this research show a power-law trend with α = 88.005 and β = 0.69 in the correlation of intensity and duration associated with antecedent cumulative rainfall (A1) anomalies. The mean return period for these anomalies resulted in 12.4 years, while for 50 % of the landslides, the recurrence interval was estimated in less than or equal to 3.6 years. In addition, significant differences were found between catalogued slope-cut failures and natural landslides. Moreover, differences were also found between simplified types of natural landslides.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have