Regional rainfall threshold maps drawn through multivariate geostatistical techniques for shallow landslide hazard zonation

Giovanna Vessia; Diego Di Curzio; Alessandro Chiaudani; Sergio Rusi

doi:10.1016/j.scitotenv.2019.135815

Regional rainfall threshold maps drawn through multivariate geostatistical techniques for shallow landslide hazard zonation

Sci Total Environ. 2020 Feb 25:705:135815. doi: 10.1016/j.scitotenv.2019.135815. Epub 2019 Nov 27.

Authors

Giovanna Vessia¹, Diego Di Curzio², Alessandro Chiaudani², Sergio Rusi²

Affiliations

¹ Department of Engineering and Geology, University "G. d'Annunzio" of Chieti-Pescara, Italy. Electronic address: g.vessia@unich.it.
² Department of Engineering and Geology, University "G. d'Annunzio" of Chieti-Pescara, Italy.

PMID: 31972946
DOI: 10.1016/j.scitotenv.2019.135815

Abstract

The Empirical Rainfall Thresholds (ERTs) for shallow landslide initiation are commonly devised worldwide mostly to be implemented within landslide early warning systems. Nonetheless, since the pioneering works on ERTs in the 1980s, only meteorological variables - that are cumulated E or intensity I and duration D values of rainfalls that are likely to trigger landslides - have been used to predict landslide occurrence, even though they are characterized by a large uncertainty. Over time, many efforts have been devoted to constrain ERT to geo-morphological characters of the landslide locations but, since nowadays, they did not get to a sound new method to derive ERT and strengthen its ability to forecast future rainfall-induced landslide. In this study, local geo-morphological characters have been taken into account by means of the co-kriging technique to constrain the E and D mean values of a regional ERT and their confidence intervals. The study area, where the proposed method was trained, is the hilly side of the Abruzzo region (Italy). Here, 62 shallow landslides have been analyzed in the time span of 2013-2017 by collecting 62 (D,E) pairs related to the rainfalls that were likely to trigger them. The relevant geo-morphological features for the considered territory have been selected through the principal component analysis. Then, the Multi-Collocated Co-Kriging technique, through ISATIS Geovariances software, has been applied to derive the spatial variability structures of E and D values conditioned by the selected geo-morphological parameters. Therefore, threshold values of E and D and their confidence intervals have been calculated generating a new shape of regional ERT, consisting of maps of continuous estimated threshold values of (D,E) and confidence interval values suitable for being used in early warning systems for shallow landslide initiation.

Keywords: Early warning systems; Multivariate geostatistics; Principal component analysis; Rainfall threshold maps; Shallow landslide triggering.