Monitoring strategies and warning models for weather-induced landslides

Abstract

Weather induced landslides cause a large number of casualties as well as severe economic losses worldwide every year. Such a diffuse risk cannot be mitigated only by means of structural works, typically characterized by significant economic and environment al impacts. Therefore, landslide early warning systems (LEWS) are being increasingly applied as non structural mitigation measures aiming at reducing the loss of life probability and other adverse consequences from landslide events by prompting people to a ct appropriately and in sufficient time to reduce the possibility of harm or loss. The systems can be distinguished, as a function of the scale of design and operation, in two different categories. Territorial systems (Te LEWS), deal with multiple landslid es over wide areas at regional scale, i.e. typically a basin, a municipality or a region; local systems (Lo LEWS) address single landslides at slope scale. In a preliminary phase of this study, a detailed review of Lo LEWS operational worldwide is provide d. The information has been retrieved from peer reviewed articles published in scientific journals and proceedings of technical conferences, books, reports, and institutional web pages. The main characteristics of these systems have been summarized and des cribed according to a scheme based on a clear distinction between three modules: landslide model, warning model and warning system. The monitoring strategies implemented therein have been presented and discussed, focusing on the monitored parameters and th e monitoring instruments for each type of landslide. Subsequently, warning models developed within Te LEWS for weather induced landslides have been analyzed , pointing out that: their outputs are strongly dependent from the accurateness and reliability of t he information on landslide occurrences; and only meteorological variables are considered in most of occurrences; and only meteorological variables are considered in most of the cases, thus leading to an unavoidable uncertainty in the empirically the cases, thus leading to an unavoidable uncertainty in the empirically defined thresholds. defined thresholds. To overcome these issues, original procedures for To overcome these issues, original procedures for defining wardefining warning models are herein proposed and tested on case studies ning models are herein proposed and tested on case studies in Campania and Emiliain Campania and Emilia--Romagna regions (Italy) and in Norway. In Italy, Romagna regions (Italy) and in Norway. In Italy, a probabilistic approach has been developed to determine landslide a probabilistic approach has been developed to determine landslide conditional probabilities related to rainfall of specific conditional probabilities related to rainfall of specific intensity and intensity and duration. The adopted Bayesian methodology allows to consider the duration. The adopted Bayesian methodology allows to consider the uncertainty of the data and to provide a quantitative assessment of the uncertainty of the data and to provide a quantitative assessment of the reliability of the results. Data on landslide occurrences have been derived reliability of the results. Data on landslide occurrences have been derived from a new landslide inventofrom a new landslide inventory, named “FraneItalia”, wherein data are ry, named “FraneItalia”, wherein data are retrieved from online journalistic news; the correlations between retrieved from online journalistic news; the correlations between landslides and rainfall have been assessed by analylandslides and rainfall have been assessed by analyzzing satelliteing satellite--rainfall rainfall records within weather alert zones. On the other hand, the methodology records within weather alert zones. On the other hand, the methodology prproposed for Norway aims at integrating the hydrooposed for Norway aims at integrating the hydro--meteorological meteorological variables employed within the regional model used by the national early variables employed within the regional model used by the national early warning system (i.e. combinations of relative water supply and relative soil warning system (i.e. combinations of relative water supply and relative soil water saturation degree) with monitoring datwater saturation degree) with monitoring data collected at local scale, a collected at local scale, specifically pore water pressure observations acquired by the Norwegian specifically pore water pressure observations acquired by the Norwegian Geotechnical Institute for a variety of projects. The analyses are carried Geotechnical Institute for a variety of projects. The analyses are carried out on a number of hydrological basins (test areas) defined at national out on a number of hydrological basins (test areas) defined at national scale andscale and selected considering the occurrence of landslides in loose soils selected considering the occurrence of landslides in loose soils from 2013 to 2017 and the availability of a significant number of pore from 2013 to 2017 and the availability of a significant number of pore water pressure measurements. For each basin, the alerts issued by the water pressure measurements. For each basin, the alerts issued by the regional model are assessed by means of a 2regional model are assessed by means of a 2--step step analysis employing analysis employing indicators derived from simple moving averages of the pore water indicators derived from simple moving averages of the pore water pressure measurements. pressure measurements. The warning models developed herein were successfully applied to The warning models developed herein were successfully applied to selected case studies. Therefore, the proposed methodologies can be selected case studies. Therefore, the proposed methodologies can be considered valuconsidered valuable frameworks considering aspects that are crucial for able frameworks considering aspects that are crucial for improving the efficiency of the models, such as: the potential of nonimproving the efficiency of the models, such as: the potential of non--conventional landslide inventories and remote sensing monitoring conventional landslide inventories and remote sensing monitoring instruments to complement the traditional sources of data, the uinstruments to complement the traditional sources of data, the use of se of probabilistic techniques for defining more objective rainfall thresholds, probabilistic techniques for defining more objective rainfall thresholds, and the additional contribution of the information derived from the local and the additional contribution of the information derived from the local observations of pore water pressures.observations of pore water pressures. [edited by Author]