D48 Case study: landslides in Azores, Faial earthquake, Furnas Volcano,1522 landslide

CVUA carried out the study of three of the most important landslide case studies that affected the Azores archipelago: (1) the 1522, 22^nd October Vila Franca do Campo debris flow; (2) the 1997, 31^st October Ribeira Quente rainfall triggered shallow slope movements; and (3) the 1998, 9^th July Faial island earthquake triggered landslides)

(1) The 1522, 22^nd October Vila Franca do Campo debris flow

The most important earthquake in the region occurred on October 22^nd, 1522. It was suggested that this event had epicentre on land, a few kilometres NNW of Vila Franca do Campo, the former capital of S. Miguel Island, reaching a maximum intensity of X (EMS-98). This earthquake induced several landslides, all over the island. The work developed in RETINA allowed to identify the geologically most important and spectacular of these landslides as cataclysmic debris flow from 450 m altitude on the southern flank of Fogo Volcano, in a place named Pico da Cruz, which buried and completely destroyed Vila Franca do Campo, located at sea level, with a loss of about 5,000 lives. This landslide involved a volume of about 6,75x10⁶ m³ and occupied an area of 4,5 km². A small tsunami has occurred, caused by the entry of the debris flow in the ocean, destroying two boats anchored near Vila Franca islet, 2 km offshore.

In order to characterize this catastrophic landslide, historical documents and aerial photographs were analysed and a detailed sedimentological and stratigraphic study was carried out. The interpretation of the available data allowed to reconstruct the main debris flow and other important debris flow deposition areas and to infer the source of each debris flows. The main debris flow deposit, as well as the other contemporaneous debris flow are well constrained in the regional stratigraphy, being covered by the 1630 Furnas Volcano eruption deposit.

The main debris flow deposit is highly variable in thickness and is very heterogeneous, considering the nature and dimension of its components. The matrix is fine-grained and mainly composed of pumice, with a significant amount of basaltic, trachytic and syenitic fragments, as well as a considerable amount of loose feldspar crystals (sanidine) and metasomatized rocks. In some outcrops, archaeological artifacts of the 16^th century were found (Fig. 10).

Fig. 10: Two of the artifacts found in the 1522 landslide deposit. a) and b) – red ceramic fragment with a large component of sandy material and a large number of muscovite crystals; c) and d) – yellow ceramic, hand painted, covered with glazing; e) – yellow ceramic covered with white glazing characteristic of Vila Franca do Campo 16^th century ceramics; f) – yellow ceramic (squares with 3 cm side).

The obtained results can be taken into account to implement strategies for land-use planning aiming to minimize the impact of future events (Fig. 11). This case study presents data of earthquake-triggered landslide in the Azores as a step for hazard assessment in the region.

Fig. 11: Flood area obtained for the 1522 debris flow case study superimposed on the 3D view of Vila Franca do Campo region.

(2) The 1997, 31^st October Ribeira Quente rainfall triggered shallow slope movements

The most recent catastrophic landslide episode took place on the 1997, 31^st October, when more than 1000 soil slips and debris flows (Fig. 12) were triggered, claiming 29 human lives in the small village of Ribeira Quente and a total loss of 21.300.000 € were spend in direct and indirect damage. The landslides were identified and characterized in WP4350, nevertheless, for this case study, the rainfall analysis was carried out using 26 years (1976/77 to 2001/02) of daily precipitation registered at the Lagoa das Furnas reference meteorological station, in order to characterize the rainfall patterns of the county.

In order to understand the antecedent rainfall influence in the catastrophe triggering, different cumulative rainfall methods were applied, to the reconstitution of cumulative rainfall, for different periods of time. These methods allowed to identify a two stage pattern of precipitation which induced the October 31^st 1997 shallow slope movements: (1) a preparatory rainfall rainy period, in the predecessor 15 days, when slopes are destabilized and conditioned for failure, followed by a more intense and short rainfall episode, in the antecedent 24 hours, that triggers the landslides. Based on the extreme values theory, which plays a basic role in small probability events, using the Gumbel extreme values distribution function, a 24,9 years return period and the occurrence probability of 4,1 % of a similar phenomena in the future were calculated.

Fig. 12: a) Typical shallow slope movement from 1997, 31^st October; b) and c) Impact on the population.

The existence of different hydrological triggering mechanisms, related to different types of landslides (deep seated or shallow slope movements) and different physical settings, makes impossible the definition of an empirically based universal rainfall threshold that can be associated with landslide occurrence. Following the same methodologies, applied in this case study, CVUA on behalf project FORESIGTH, will develop a model for rainfall-triggered landslides, based on patterns and rainfall critical values for landslide occurrence. Considering the critical pairs rainfall amount-duration for each landslide occurrence, we will try to define a rainfall-triggering threshold. The identification of rainfall patterns associated to landslide events will lead to the development of prevention, prediction, alert and warning systems for landslide risk mitigation.

(3) The 1998, 9^th July Faial Island earthquake triggered landslides.

For this case study, the landslides, mainly debris flows or shallow landslides that commonly mobilize into debris flows (Fig. 13), were identified and mapped using aerial photo and field surveys, carried out on July 1998.

Fig. 13: a) Shallow slope movements inside the Faial caldera; b) Debris flow in the flank of the volcano.

In the future we will use traditional landslide hazard mapping approaches, combining basic input variables such as lithology, vegetation and morphometric characteristics (slope, curvature, etc.) that will be used in order to calculate the probability of landslide occurrence. Soil Mechanical Response Units (SMRU’s) will be delimited using a Geographical Information System (GIS) combining geological and lithological characteristics. The cohesion and internal friction angle, for each SMRU will be derived from standard values from other works obtained by shear strength tests. These will be used in combination with morphometric parameters, derived from a digital elevation model. The Factors of Safety for each considered cell will be calculated using the infinite slope model, a 2D model describing the stability of slopes with an infinitively large failure plane. The disruption caused by an earthquake will be computed using the earthquake peak acceleration. In order to validate this methodology, it will be computed the effect of the 1998 Faial earthquake in the Factor of Safety using the seismic acceleration peak. The result will be compared to the localization earthquake triggered landslides identified on behalf project RETINA. The final product of this investigation will be a landslide hazard map that can be used in future land-use policy changes that need to be made to mitigate the hazard derived from landslide risk.

This work package has not been completed. There are two main reasons for this: 1) The seismic and ground water flow model of WP 3400 is not yet capable of dealing with complex 3D situations such as presented by the geology and topography of Furnas and Fogo volcanoes; and 2) The topographic data required for full characterization of the volcanoes doesn’t exist yet. The existing topographic data that belongs to the Cartographic Services of Portuguese Army has contour intervals of 10 metres which is not sufficient for a good application of the models.

However, UCAM.DES has successfully modelled the response of cone-shaped topographies to seismic perturbations. The key features of this response are likely to apply to volcanoes such as Fogo and Furnas. Modelling of more realistic volcano cross-sections will follow.

Modelling by UCAM.DES will continue in FORESIGHT project (FP6 project). Inspiration will be drawn from historic events in the Azores.