EU6-Italy

HOME

In Italy, intense or prolonged rainfall is the primary trigger of landslides. Damaging failures occur every year in this country, where they have caused more than 6300 casualties in the 60-year period, 1950–2009 (Salvati et al., 2010). the Italian National Department for Civil Protection (DPC), an Office of the Prime Minister, asked the Research Institute for Geo-Hydrological Hazard Assessment (IRPI), of the Italian National Research Council (CNR), to design and implement an early-warning system to forecast the possible occurrence of rainfall- induced landslides in Italy. The Italian CNR,IRPI designed and implemented a landslide warning system, named SANF (an acronym for national early warning system for rainfall-induced landslides), to forecast the possible occurrence of rainfall-induced landslides in Italy. The SANF early warning system for the forecast of the possible occurrence of rainfall-induced landslides in Italy has been operational since October 2009.

1 The setting

The system is based on: (i) rainfall thresholds for possible landslide occurrences, (ii) sub-hourly rainfall measurements obtained by a national network of 1950 rain gauges, and (iii) quantitative rainfall forecasts. Twice a day, the system compares the measured and the forecasted rainfall amounts against pre-defined ID thresholds, and assigns to each rain gauge a probability of landslide occurrence. This information is used to prepare synoptic-scale maps showing where rainfall-induced landslides are expected in the next 24 hours.

2 The modelling

Empirical thresholds are obtained statistically from studying past rainfall events that have resulted in slope failures. For each landslide event in the database, the rainfall duration (D) and the rainfall mean intensity (I) that have resulted in the slope instability are established analyzing the rainfall record of the most representative rain gauge. For most of the landslides, the representative rain gauge was the closest to the landslide. To define reproducible, objective and reliable thresholds for possible landslide occurrence, IRPI has devised a specific method. The method assumes that the threshold curve is a power law, I = α⋅D−β, where, I, is the mean rainfall intensity (in mm/h), D is the rainfall duration (in h), and α and β are positive coefficients. Currently, the system uses a single threshold with 1% exceedance probability defined for the entire Italian territory (Fig. 21).

Figure 21: Intensity-duration conditions (dots) that resulted landslides in Italy. Black line is the rainfall threshold at 1% exceedance probability implemented in the SANF early warning system.

3 The warning strategy

For each rainfall event, the difference between the event intensity and the intensity of the fitted power law is calculated. The probability density of the distribution of the differences is determined through Kernel Density Estimation, and the result fitted with a Gaussian function. Finally, thresholds corresponding to different exceedance probabilities are defined (Brunetti et al., 2009). The scheme (Fig. 18) is based on four Frequentist thresholds, namely: T0.005 - T0.5- T1.5 -T5, corresponding to an exceedance probability of 0.005%, 0.5%, 1.5% and 5% of the area under the Gaussian fit (Fig. 22). In the scheme, the four thresholds separate five ID fields (shown by different colors in Fig. 18). For any given rainfall duration, D, when the (measured or predicted) rainfall mean intensity, I, is lower than the 0.005% threshold, the rainfall condition is considered “well below the threshold” (level 1). Similarly, when the rainfall mean-intensity, I, is between the 0.005% and the 0.5% thresholds, the rainfall condition is considered “below the threshold” (level 2). When the rainfall mean intensity, I, is in the range between the 0.5% and the 1.5% thresholds or in the range between the 1.5% and the 5% thresholds, the rainfall condition is considered “on the threshold” (level 3) or “above the threshold” (level 4), respectively. Lastly, when the rainfall mean intensity, I, is equal to, or larger than, the upper 5% threshold, the rainfall condition is considered “well above the threshold” (level 5). In this area, landslides are typically expected, with a chance of false negatives of 5.0%, or more.

Figure 22: Critical rainfall conditions defined by thresholds having different exceedance probability shown (a) in the Gaussian curve (see Fig. 2), and (b) in the D-I plane. Legend: dark green, rainfall condition “well below the threshold”; light green, “below the threshold”; yellow, “on the threshold”; orange, “above the threshold”; red, “well above the threshold”.

3 The warning model

Input data for SANF include: (i) rainfall measurements, (ii) quantitative rainfall forecasts, and (iii) information on past rainfall events that have resulted in landslides in Italy (Fig. 23). A network of 1950 rain gauges provides rainfall measurements in Italy. Every six hours, rainfall measurements from this network are imported, validated, and stored in the system. Three-day quantitative rainfall forecasts (Fig. 3) are generated twice a day by DPC using the Italian Local Area Model (LAMI), and stored in the system.

Every 12 hours, for each rain gauge, critical levels for the available rainfall measurements and rainfall forecasts are calculated. The critical levels for the rainfall measurements are calculated using the antecedent rainfall measured at durations of 24, 48, 72, and 96 hours (Fig. 24.a). Levels obtained from the comparison of the measured rainfall (for each duration) and the corresponding values on the rainfall thresholds (L−24, L−48, L−72, and L−96) are combined to calculate the resulting critical level (Ls). If the level obtained for the antecedent 24-hour rainfall (L−24) is larger than the other levels, then the L−24 is taken as the critical level. Else, the critical level is obtained through a weighted mean of the four levels.

Figure 24: Example of rainfall intensity-duration conditions used for the calculation of rain gauges critical levels (Ls) using rainfall measures (A) and rainfall forecasts (B). (a) Periods considered for the calculation of Ls using rainfall measures. (b) Periods considered for the calculation of Ls using rainfall forecasts.

Figure 23: Logical framework adopted for the SANF.

In Equation (1), the weights (Wi) decrease linearly with the duration of the considered antecedent periods. The rain gauge critical level for rainfall forecasts is obtained in a similar way, using the forecasted rainfall at 24 and 48 hours (Fig. 24.b).

The National Department for Civil Protection (DPC) has subdivided the Italian territory in 129 alert zones. Critical levels of rain gauges belonging to the same alert zone are aggregated to define alert zone critical levels. The system uses two different aggregation criteria, based on the maximum and the modal values, respectively. Following the first criteria, the system assigns to the alert zone the highest level among those reached by rain gauges in the alert zone. Using the second criteria, the system assigns to the alert zone the most frequent (modal) level among those reached by rain gauges in the zone.

The early warning system computes critical levels for each rain gauge and for each alert zone every 12 hours. When the critical levels are determined, the system generates synoptic-scale maps of critical levels for the possible occurrence of rainfall-induced landslides in Italy. The maps (i.e., the services) can be consulted using both open source and proprietary software (e.g., Quantum GIS, ESRI ArcMap). For convenience, the services were integrated in a dedicated WebGIS interface.

Once a day, a specific procedure generates a summary report (in PDF format) containing, for both rainfall measurements and rainfall forecasts, a map showing the rain gauge critical levels, and two maps showing the alert zone critical levels (maximum and modal). The report is delivered via e-mail to the Italian National Department for Civil Protection.

The rainfall ID threshold currently used in the system was validated using a reduced set of independent landslide information for a six-month period in the Abruzzo region.

References

M. Rossi, S. Peruccacci, M.T. Brunetti, I. Marchesini, S. Luciani, F. Ardizzone, V. Balducci, C. Bianchi, M. Cardinali, F. Fiorucci, A.C. Mondini, P. Reichenbach, P. Salvati, M. Santangelo, D. Bartolini, S.L. Gariano, M. Palladino, G. Vessia, A. Viero, L. Antronico, L. Borselli, A.M. Deganutti, G. Iovine, F. Luino, M. Parise, M. Polemio & F. Guzzetti, S. Luciani, F. Fiorucci, A.C. Mondini & M. Santangelo, G. Tonelli, (2012). SANF: National warning system for rainfall-induced landslides in Italy. Landslides and Engineered Slopes: Protecting Society through Improved Understanding – Eberhardt et al. (eds) © 2012 Taylor & Francis Group, London, ISBN 978-0-415-62123-6

Brunetti, M.T., Peruccacci, S., Rossi, M., Luciani, S., Valigi, D. & Guzzetti, F. 2010. Rainfall thresholds for the possible occurrence of landslides in Italy. Natural Hazards and Earth System Sciences 10: 447–458.