Rainfall-Induced Landslides forecast using local precipitation and global climate indexes

• Published in: Natural Hazards
• Main Authors: Fustos, I, Abarca del Rio, R., Moreno Yaeger, P., Somos Valenzuela, M.
• Format: Article in Journal/Newspaper
• Language: English
• Published: SPRINGER 2021
• Subjects: Rainfall-Induced Landslides; logistic regression; ENSO-AAO variability; Antarctic; Pacific; The Antarctic; Antarc*
• Online Access: http://repositoriodigital.uct.cl/handle/10925/3897; https://doi.org/10.1007/s11069-020-03913-0

We analyse RIL events between 1950 and 2002 to investigate the role played by climate variability, using the 'El Nino-Southern Oscillation' (ENSO), the Antarctic Oscillation (AAO) and local precipitation as predictors, through logistic and probabilistic (Logit and Probit) modelling. From the probabilistic regression analysis, it is clear that rain plays a major role, since its weight in the regression is almost 50%. However, we show that integrating South Pacific climate variability represented by ENSO/AAO significantly increases predictability, reaching over 87%. Moreover, sensitivity and specificity analyses confirm that although local rainfall is the main triggering factor, adding the two macroclimate variables increases the ability to predict true positive and negative occurrences by almost 80%. This confirms the need to integrate macroclimatic variables to make assertive local predictions. Surprisingly, and contrary to what might have been expected considering ENSO's recognized role in regional climate variability, the integration of AAO variability significantly improves RIL prediction capacity, while on average ENSO can be considered a second-order predictor. These results, obtained through a simple logistic regression methodology (Logit and/or Probit), can contribute to better risk management in the middle-latitude zones of Chile. The methodology can be extended to other areas of the world that do not have high-density hydrometeorological information to support preventive decision-making through logistic RIL forecasting.